| blob | 8a848124ff732e418b0fa6a39c9273b285cfc90c |
1 2018-05-16 Martin Jambor <mjambor@suse.cz>
3 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
8 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
9 (fnma<mode>4): Likewise.
10 (fms<mode>4): Likewise.
11 (fnms<mode>4): Likewise.
12 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13 (aarch64_fnma<mode>4): Likewise.
14 (aarch64_fms<mode>4): Likewise.
15 (aarch64_fnms<mode>4): Likewise.
16 (aarch64_fnmadd<mode>4): Likewise.
18 2018-05-16 Jason Merrill <jason@redhat.com>
20 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
22 2018-05-16 Richard Biener <rguenther@suse.de>
24 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
25 (dump_stmt_cost): Declare.
26 (add_stmt_cost): Dump cost we add.
27 (add_stmt_costs): New function.
28 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
29 No longer exported.
30 (vect_analyze_stmt): Adjust prototype.
31 (vectorizable_condition): Likewise.
32 (vectorizable_live_operation): Likewise.
33 (vectorizable_reduction): Likewise.
34 (vectorizable_induction): Likewise.
35 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
36 cost vector to pass to vectorizable_ and record afterwards.
37 (vect_model_reduction_cost): Take cost vector argument and adjust.
38 (vect_model_induction_cost): Likewise.
39 (vectorizable_reduction): Likewise.
40 (vectorizable_induction): Likewise.
41 (vectorizable_live_operation): Likewise.
42 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
43 SLP_TREE_NUMBER_OF_VEC_STMTS.
44 (vect_analyze_slp_cost_1): Remove.
45 (vect_analyze_slp_cost): Likewise.
46 (vect_slp_analyze_node_operations): Take visited args and
47 a target cost vector. Avoid processing already visited stmt sets.
48 (vect_slp_analyze_operations): Use a local cost vector to gather
49 costs and register those of non-discarded instances.
50 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
51 (vect_schedule_slp_instance): Remove copying of
52 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
53 zero.
54 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
55 adding cost. Record cost entry location.
56 (vect_prologue_cost_for_slp_op): Function to compute cost of
57 a constant or invariant generated for SLP vect in the prologue,
58 split out from vect_analyze_slp_cost_1.
59 (vect_model_simple_cost): Make static. Adjust for SLP costing.
60 (vect_model_promotion_demotion_cost): Likewise.
61 (vect_model_store_cost): Likewise, make static.
62 (vect_model_load_cost): Likewise.
63 (vectorizable_bswap): Add cost vector arg and adjust.
64 (vectorizable_call): Likewise.
65 (vectorizable_simd_clone_call): Likewise.
66 (vectorizable_conversion): Likewise.
67 (vectorizable_assignment): Likewise.
68 (vectorizable_shift): Likewise.
69 (vectorizable_operation): Likewise.
70 (vectorizable_store): Likewise.
71 (vectorizable_load): Likewise.
72 (vectorizable_condition): Likewise.
73 (vectorizable_comparison): Likewise.
74 (can_vectorize_live_stmts): Likewise.
75 (vect_analyze_stmt): Likewise.
76 (vect_transform_stmt): Adjust calls to vectorizable_*.
77 * tree-vectorizer.c: Include gimple-pretty-print.h.
78 (dump_stmt_cost): New function.
80 2018-05-16 Richard Biener <rguenther@suse.de>
82 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
83 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
84 * tree-ssa-dse.c: Include tree-ssa-loop.h.
85 (check_name): New callback.
86 (dse_classify_store): Track cycles via a visited bitmap of PHI
87 defs and simplify handling of in-loop and across loop dead stores
88 and properly fail for loop-variant refs. Handle byte-tracking with
89 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
90 limiting the walk.
92 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
94 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
95 (vect_get_mask_type_for_stmt): Likewise.
96 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
97 split out from...
98 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
99 to determine the statement's vector type and the vector type that
100 should be used for calculating nunits. Deal with cases in which
101 the type has to be deferred.
102 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
103 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
104 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
105 (vect_determine_vf_for_stmt): New functions, split out from...
106 (vect_determine_vectorization_factor): ...here.
107 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
108 (vect_get_mask_type_for_stmt): New functions, split out from
109 vect_determine_vectorization_factor.
111 2018-05-16 Richard Biener <rguenther@suse.de>
113 * tree-cfg.c (verify_gimple_assign_ternary): Properly
114 verify the [VEC_]COND_EXPR embedded comparison.
116 2018-05-15 Martin Sebor <msebor@redhat.com>
118 PR tree-optimization/85753
119 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
120 RECORD_TYPE in addition to ARRAY_TYPE.
122 2018-05-15 Martin Sebor <msebor@redhat.com>
124 PR middle-end/85643
125 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
127 2018-05-15 Richard Biener <rguenther@suse.de>
129 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
130 add by_clobber_p one. Change algorithm to collect all defs
131 representing uses we need to walk and try reducing them to
132 a single one before failing.
133 (dse_dom_walker::dse_optimize_stmt): Adjust.
135 2018-05-13 Mark Wielaard <mark@klomp.org>
137 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
138 (size_of_loc_descr): Likewise.
139 (output_loc_operands): Likewise.
140 (output_loc_operands_raw): Likewise.
141 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
142 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
143 (hash_loc_operands): Likewise.
144 (compare_loc_operands): Likewise.
146 2018-05-14 Mark Wielaard <mark@klomp.org>
148 * dwarf2out.c (count_index_addrs): New function.
149 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
151 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
153 PR tree-optimization/83648
154 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
155 return value as malloc candidate.
157 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
159 PR ipa/85734
160 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
161 as true in call to suggest_attribute.
163 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
165 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
166 -mreadonly-in-sdata.
168 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
170 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
171 New pattern.
172 (aarch64_crypto_aesd_fused): Likewise.
174 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
176 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
177 (movsi_aarch64): Likewise.
178 (load_pairsi): Likewise.
179 (load_pairdi): Likewise.
180 (store_pairsi): Likewise.
181 (store_pairdi): Likewise.
182 (load_pairsf): Likewise.
183 (load_pairdf): Likewise.
184 (store_pairsf): Likewise.
185 (store_pairdf): Likewise.
186 (zero_extend): Likewise.
187 (trunc): Swap alternatives.
188 (fcvt_target): Add '?' to prefer w over r.
190 2018-05-14 Jakub Jelinek <jakub@redhat.com>
192 PR target/85756
193 * config/i386/i386.md: Disallow non-commutative arithmetics in
194 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
195 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
196 in the peephole2 before it.
198 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
200 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
201 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
202 (ix86_handle_option): Handle -mcldemote.
203 * config.gcc: New header.
204 * config/i386/cldemoteintrin.h: New file.
205 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
206 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
207 -mcldemote.
208 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
209 OPTION_MASK_ISA_CLDEMOTE.
210 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
211 (ix86_valid_target_attribute_inner_p): Ditto.
212 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
213 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
214 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
215 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
216 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
217 (cldemote): New.
218 * config/i386/i386.opt: Add -mcldemote.
219 * config/i386/x86intrin.h: New header.
220 * doc/invoke.texi: Add -mcldemote.
222 2018-05-14 Richard Biener <rguenther@suse.de>
224 * doc/match-and-simplify.texi: Adjust :s documentation.
226 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
228 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
229 intended memcpy size.
230 (REORDER_45): Likewise.
232 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
234 * sort.cc: New file.
235 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
236 * vec.c (qsort_chk): Use gcc_qsort.
237 * Makefile.in (OBJS-libcommon): Add sort.o.
238 (build/sort.o): New target. Use it...
239 (BUILD_RTL): ... here, and...
240 (build/gencfn-macros): ... here, and...
241 (build/genmatch): ... here.
243 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
244 Chung-Ju Wu <jasonwucj@gmail.com>
246 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
247 * config/nds32/nds32-graywolf.md: New file.
248 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
249 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
250 pipeline.
251 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
252 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
253 * config/nds32/nds32.md (pipeline_model): Add graywolf.
254 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
255 * config/nds32/pipelines.md: Include n15 settings.
257 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
258 Chung-Ju Wu <jasonwucj@gmail.com>
260 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
261 * config/nds32/nds32-n13.md: New file.
262 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
263 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
264 pipeline.
265 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
266 * config/nds32/nds32.md (pipeline_model): Add n13.
267 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
268 * config/nds32/pipelines.md: Include n13 settings.
270 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
271 Chung-Ju Wu <jasonwucj@gmail.com>
273 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
274 * config/nds32/nds32-n10.md: New file.
275 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
276 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
277 pipeline.
278 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
279 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
280 * config/nds32/nds32.md (pipeline_model): Add n10.
281 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
282 * config/nds32/pipelines.md: Include n10 settings.
284 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
285 Kito Cheng <kito.cheng@gmail.com>
286 Chung-Ju Wu <jasonwucj@gmail.com>
288 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
289 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
290 Add enum values for DSP extension instructions.
291 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
292 New constraints.
293 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
294 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
295 New code iterators.
296 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
297 * config/nds32/nds32-dspext.md: New file for DSP implementation.
298 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
299 * config/nds32/nds32-intrinsic.md: Likewise.
300 * config/nds32/nds32_intrinsic.h: Likewise.
301 * config/nds32/nds32-md-auxiliary.c: Likewise.
302 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
303 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
304 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
305 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
306 * config/nds32/nds32-protos.h: New declarations for DSP extension.
307 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
308 TYPE_DMAC in switch statement.
309 * config/nds32/nds32.c: New checking and implementation for DSP
310 extension instructions.
311 * config/nds32/nds32.h: Likewise.
312 * config/nds32/nds32.md: Likewise.
313 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
314 * config/nds32/predicates.md: Implement new predicates for DSP
315 extension.
317 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
319 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
320 Reformat alternatives and attributes so it is easier to identify
321 which constraints/attributes go with which instruction.
322 (mov<mode>_hardfloat32, FMOVE64): Likewise.
323 (mov<mode>_softfloat32, FMOVE64): Likewise.
324 (mov<mode>_hardfloat64, FMOVE64): Likewise.
325 (mov<mode>_softfloat64, FMOVE64): Likewise.
327 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
329 * doc/extend.texi (PowerPC Built-in Functions): Rename this
330 subsection.
331 (Basic PowerPC Built-in Functions): The new name of the
332 subsection previously known as "PowerPC Built-in Functions".
333 (Basic PowerPC Built-in Functions Available on all Configurations):
334 New subsubsection.
335 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
336 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
337 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
338 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
340 2018-05-11 Martin Jambor <mjambor@suse.cz>
342 PR ipa/85655
343 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
344 single const.
346 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
348 PR target/85733
349 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
351 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
353 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
354 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
355 (ix86_handle_option): Handle -mwaitpkg.
356 * config.gcc: New header.
357 * config/i386/cpuid.h (bit_WAITPKG): New bit.
358 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
359 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
360 function type.
361 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
362 OPTION_MASK_ISA_WAITPKG.
363 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
364 (ix86_option_override_internal): Add PTA_WAITPKG.
365 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
366 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
367 IX86_BUILTIN_TPAUSE.
368 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
369 __builtin_ia32_umwait and __builtin_ia32_tpause.
370 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
371 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
372 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
373 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
374 UNSPECV_TPAUSE): New.
375 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
376 * config/i386/i386.opt: Add -mwaitpkg.
377 * config/i386/waitpkgintrin.h: New file.
378 * config/i386/x86intrin.h: New header.
379 * doc/invoke.texi: Add -mwaitpkg.
381 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
383 PR target/85606
384 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
385 equivalent.
386 (cortex-m0): Use armv6s-m isa.
387 (cortex-m0plus): Likewise.
388 (cortex-m1): Likewise.
389 (cortex-m0.small-multiply): Likewise.
390 (cortex-m0plus.small-multiply): Likewise.
391 (cortex-m1.small-multiply): Likewise.
393 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
394 Jakub Jelinek <jakub@redhat.com>
396 PR tree-optimization/85692
397 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
398 source permute as well.
400 2018-05-11 Martin Liska <mliska@suse.cz>
402 PR sanitizer/85556
403 * doc/extend.texi: Document LLVM style format for no_sanitize
404 attribute.
406 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
408 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
409 mode_supports_vsx_dform_quad to mode_supports_dq_form.
410 (mode_supports_vsx_dform_quad): Likewise.
411 (mode_supports_vmx_dform): Move these functions to be next to the
412 other mode_supports functions.
413 (mode_supports_dq_form): Likewise.
414 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
415 mode_supports_dq_form.
416 (reg_offset_addressing_ok_p): Likewise.
417 (offsettable_ok_by_alignment): Likewise.
418 (rs6000_legitimate_offset_address_p): Likewise.
419 (legitimate_lo_sum_address_p): Likewise.
420 (rs6000_legitimize_address): Likewise.
421 (rs6000_legitimize_reload_address): Likewise.
422 (rs6000_secondary_reload_inner): Likewise.
423 (rs6000_preferred_reload_class): Likewise.
424 (rs6000_output_move_128bit): Likewise.
426 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
428 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
429 Generate SImode target register for null target.
430 <case IX86_BUILTIN_XGETBV>: Ditto.
431 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
432 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
434 2018-05-10 Carl Love <cel@us.ibm.com>
436 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
437 dcbtt and dcbtstt if operands[2] is 0.
439 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
441 PR target/85693
442 * config/i386/sse.md (usadv64qi): New expander.
444 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
446 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
447 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
448 -maltivec=be support.
449 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
450 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
451 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
452 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
453 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
454 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
455 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
456 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
457 altivec_vsumsws): Adjust.
458 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
459 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
460 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
461 support.
462 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
463 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
464 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
465 (altivec_lve<VI_char>x): Delete expand.
466 (*altivec_lve<VI_char>x_internal): Rename to...
467 (altivec_lve<VI_char>x): ... this.
468 (altivec_lvxl_<mode>): Delete expand.
469 (*altivec_lvxl_<mode>_internal): Rename to ...
470 (altivec_lvxl_<mode>): ... this.
471 (altivec_stvxl_<mode>): Delete expand.
472 (*altivec_stvxl_<mode>_internal): Rename to ...
473 (altivec_stvxl_<mode>): ... this.
474 (altivec_stve<VI_char>x): Delete expand.
475 (*altivec_stve<VI_char>x_internal): Rename to ...
476 (altivec_stve<VI_char>x): ... this.
477 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
478 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
479 reduc_plus_scal_<mode>): Adjust.
480 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
481 comment.
482 (rs6000_cpu_cpp_builtins): Adjust.
483 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
484 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
485 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
486 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
487 -maltivec=be support.
488 (rs6000_split_vec_extract_var): Adjust.
489 (rs6000_split_v4si_init): Adjust.
490 (swap_selector_for_mode): Delete.
491 (altivec_expand_lvx_be, altivec_expand_stvx_be,
492 altivec_expand_stvex_be): Delete.
493 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
494 -maltivec=be support.
495 (rs6000_gimple_fold_builtin): Ditto.
496 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
497 Adjust.
498 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
499 (TARGET_DIRECT_MOVE_64BIT): Adjust.
500 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
501 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
502 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
503 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
504 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
505 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
506 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
507 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
508 anonymous split): Adjust.
509 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
510 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
512 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
514 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
515 when --with-gxx-include-dir is also specified.
516 * configure: Regenerate.
518 2018-05-09 Jim Wilson <jimw@sifive.com>
520 PR target/84797
521 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
522 * config/riscv/t-withmultilib: New.
523 * config/riscv/withmultilib.h: New.
524 * doc/install.texi: Document RISC-V --with-multilib-list support.
526 2018-05-09 Richard Biener <rguenther@suse.de>
528 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
529 vector.
530 (vect_bb_vectorization_profitable_p): Adjust. Compute
531 actual scalar cost using the cost vector and the add_stmt_cost
532 machinery.
534 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
536 PR rtl-optimization/85645
537 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
538 in the REG_CFA_REGISTER note for LR, don't leave it empty.
540 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
542 PR rtl-optimization/85645
543 * shrink-wrap.c (spread_components): Return a boolean saying if
544 anything was changed.
545 (try_shrink_wrapping_separate): Iterate spread_components until
546 nothing changes anymore.
548 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
550 PR rtl-optimization/85645
551 * regrename.c (build_def_use): Also kill the chains that include the
552 destination of a REG_CFA_REGISTER note.
554 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
556 PR rtl-optimization/85645
557 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
558 insn that has a REG_CFA_REGISTER note.
560 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
562 * cfgexpand.c (expand_clobber): New function.
563 (expand_gimple_stmt_1): Use it.
564 * tree-vect-stmts.c (vect_clobber_variable): New function,
565 split out from...
566 (vectorizable_simd_clone_call): ...here.
567 (vectorizable_store): Emit a clobber either side of an
568 IFN_STORE_LANES sequence.
569 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
571 2018-05-09 Tom de Vries <tom@codesourcery.com>
573 PR target/85626
574 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
575 (define_insn "trap_if_false"): Add exit after trap.
577 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
579 PR rtl-optimization/85638
580 * bb-reorder.c: Include common/common-target.h.
581 (create_forwarder_block): New function extracted from...
582 (fix_up_crossing_landing_pad): ...here. Rename into...
583 (dw2_fix_up_crossing_landing_pad): ...this.
584 (sjlj_fix_up_crossing_landing_pad): New function.
585 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
586 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
587 from both partitions and exit the loop after one iteration.
589 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
591 Revert:
592 * doc/extend.texi (PowerPC Built-in Functions): Rename this
593 subsection.
594 (Basic PowerPC Built-in Functions): The new name of the
595 subsection previously known as "PowerPC Built-in Functions".
596 (Basic PowerPC Built-in Functions Available on all Configurations):
597 New subsubsection.
598 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
599 subsubsection.
600 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
601 subsubsection.
602 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
603 subsubsection.
604 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
605 subsubsection.
607 2018-05-08 Jim Wilson <jimw@sifive.com>
609 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
610 (LD_EMUL_SUFFIX): New.
611 (LINK_SPEC): Use it.
613 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
615 * doc/extend.texi (PowerPC Built-in Functions): Rename this
616 subsection.
617 (Basic PowerPC Built-in Functions): The new name of the
618 subsection previously known as "PowerPC Built-in Functions".
619 (Basic PowerPC Built-in Functions Available on all Configurations):
620 New subsubsection.
621 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
622 subsubsection.
623 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
624 subsubsection.
625 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
626 subsubsection.
627 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
628 subsubsection.
630 2018-05-08 Jakub Jelinek <jakub@redhat.com>
632 PR target/85683
633 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
634 after cmpelim optimization.
636 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
638 * config.gcc: Support "goldmont".
639 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
640 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
641 PROCESSOR_GOLDMONT.
642 * config/i386/i386.c (m_GOLDMONT): Define.
643 (processor_target_table): Add "goldmont".
644 (PTA_GOLDMONT): Define.
645 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
646 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
647 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
648 (fold_builtin_cpu): Add "goldmont".
649 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
650 (ix86_option_override_internal): Add "goldmont".
651 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
652 (processor_type): Add PROCESSOR_GOLDMONT.
653 * config/i386/i386.md: Add CPU "glm".
654 * config/i386/glm.md: New file.
655 * config/i386/x86-tune.def: Add m_GOLDMONT.
656 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
658 2018-05-08 Jakub Jelinek <jakub@redhat.com>
660 PR target/85572
661 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
662 E_V4DImode.
663 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
664 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
665 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
667 PR target/85317
668 * config/i386/i386.c (ix86_fold_builtin): Handle
669 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
671 PR target/85480
672 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
673 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
675 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
677 PR target/85658
678 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
679 (check_arch): Likewise.
680 (check_fpu): Return the result rather than printing it.
681 (end arch): Fix operator precedence.
682 (end cpu): Likewise.
683 (END): Print the result from check_fpu.
685 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
686 Alan Hayward <alan.hayward@arm.com>
687 David Sherwood <david.sherwood@arm.com>
689 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
690 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
691 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
692 (*fcmuo<mode>_and): New patterns.
694 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
696 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
697 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
698 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
699 (cmp_op, sve_imm_con): New code attributes.
700 (SVE_COND_INT_CMP, imm_con): Delete.
701 (cmp_op): Remove above unspecs from int attribute.
702 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
703 to...
704 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
705 comparison-specific unspecs.
706 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
707 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
708 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
709 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
710 (*vec_fcm<cmp_op><mode>): Rename to...
711 (*fcm<cmp_op><mode>): ...this and adjust likewise.
712 (*vec_fcmuo<mode>): Rename to...
713 (*fcmuo<mode>): ...this and adjust likewise.
714 (*pred_fcm<cmp_op><mode>): New pattern.
715 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
716 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
717 functions.
718 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
719 and UNORDERED.
720 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
721 (aarch64_emit_sve_predicated_cond): New function.
722 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
723 (aarch64_emit_unspec_cond_or): Replace with...
724 (aarch64_emit_sve_or_conds): ...this new function. Use
725 aarch64_emit_sve_ptrue_op for the individual comparisons and
726 aarch64_emit_binop to OR them together.
727 (aarch64_emit_inverted_unspec_cond): Replace with...
728 (aarch64_emit_sve_inverted_cond): ...this new function. Use
729 aarch64_emit_sve_ptrue_op for the comparison and
730 aarch64_emit_unop to invert the result.
731 (aarch64_expand_sve_vec_cmp_float): Update after the above
732 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
734 2018-05-07 Nathan Sidwell <nathan@acm.org>
736 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
737 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
738 (Backwards Compatibility): Likewise.
740 2018-05-07 Luis Machado <luis.machado@linaro.org>
742 PR bootstrap/85681
743 Revert:
744 2018-05-07 Luis Machado <luis.machado@linaro.org>
746 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
747 <prefetch_dynamic_strides>: New const bool field.
748 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
749 prefetch_dynamic_strides.
750 (exynosm1_prefetch_tune): Likewise.
751 (thunderxt88_prefetch_tune): Likewise.
752 (thunderx_prefetch_tune): Likewise.
753 (thunderx2t99_prefetch_tune): Likewise.
754 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
755 (aarch64_override_options_internal): Update to set
756 PARAM_PREFETCH_DYNAMIC_STRIDES.
757 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
758 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
759 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
760 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
761 prefetch-dynamic-strides setting.
763 2018-05-07 Luis Machado <luis.machado@linaro.org>
765 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
766 <minimum_stride>: New const int field.
767 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
768 minimum_stride field.
769 (exynosm1_prefetch_tune): Likewise.
770 (thunderxt88_prefetch_tune): Likewise.
771 (thunderx_prefetch_tune): Likewise.
772 (thunderx2t99_prefetch_tune): Likewise.
773 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
774 (aarch64_override_options_internal): Update to set
775 PARAM_PREFETCH_MINIMUM_STRIDE.
776 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
777 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
778 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
779 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
780 stride is constant and is below the minimum stride threshold.
782 2018-05-07 Luis Machado <luis.machado@linaro.org>
784 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
785 to 512.
787 2018-05-07 Luis Machado <luis.machado@linaro.org>
789 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
790 <prefetch_dynamic_strides>: New const bool field.
791 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
792 prefetch_dynamic_strides.
793 (exynosm1_prefetch_tune): Likewise.
794 (thunderxt88_prefetch_tune): Likewise.
795 (thunderx_prefetch_tune): Likewise.
796 (thunderx2t99_prefetch_tune): Likewise.
797 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
798 (aarch64_override_options_internal): Update to set
799 PARAM_PREFETCH_DYNAMIC_STRIDES.
800 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
801 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
802 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
803 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
804 prefetch-dynamic-strides setting.
806 2018-05-07 Luis Machado <luis.machado@linaro.org>
808 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
809 <minimum_stride>: New const int field.
810 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
811 minimum_stride field.
812 (exynosm1_prefetch_tune): Likewise.
813 (thunderxt88_prefetch_tune): Likewise.
814 (thunderx_prefetch_tune): Likewise.
815 (thunderx2t99_prefetch_tune): Likewise.
816 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
817 (aarch64_override_options_internal): Update to set
818 PARAM_PREFETCH_MINIMUM_STRIDE.
819 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
820 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
821 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
822 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
823 stride is constant and is below the minimum stride threshold.
825 2018-05-06 Jakub Jelinek <jakub@redhat.com>
827 PR c++/85659
828 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
829 the type is addressable. Don't force op into register if it has
830 BLKmode.
832 2018-05-05 Roland McGrath <mcgrathr@google.com>
834 PR other/77609
835 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
836 any section for which we don't know a specific type it should have,
837 regardless of name. Previously this was done only for the exact
838 names ".init_array", ".fini_array", and ".preinit_array".
839 (default_elf_asm_named_section): Add comment about
840 relationship with default_section_type_flags and SECTION_NOTYPE.
841 (get_section): Don't consider it a type conflict if one side has
842 SECTION_NOTYPE and the other doesn't, as long as neither has the
843 SECTION_BSS et al used in the default_section_type_flags logic.
845 2018-05-05 Tom de Vries <tom@codesourcery.com>
847 PR target/85653
848 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
849 (workaround_barsyncs): New function.
850 (nvptx_reorg): Use workaround_barsyncs.
851 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
852 (define_expand "nvptx_membar_cta"): New define_expand.
853 (define_insn "*nvptx_membar_cta"): New insn.
855 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
857 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
858 To improve optimization opportunities.
859 * builtin-types.def: The new needed builtin types for the above.
861 2018-05-04 Richard Biener <rguenther@suse.de>
863 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
864 * gimple-ssa-store-merging.c
865 (imm_store_chain_info::output_merged_store): Remove redundant create,
866 release split_store vector contents on failure.
867 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
868 scalar stmt vector on cache hit.
870 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
872 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
873 Xilinx FP support.
874 * config.gcc (powerpc-xilinx-eabi*): Remove.
875 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
876 support.
877 (fusion_addis_mem_combo_load): Ditto.
878 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
879 FP support.
880 (rs6000_cpu_cpp_builtins): Ditto.
881 * config/rs6000/rs6000-linux.c
882 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
883 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
884 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
885 support.
886 (rs6000_setup_reg_addr_masks): Ditto.
887 (rs6000_init_hard_regno_mode_ok): Ditto.
888 (rs6000_option_override_internal): Ditto.
889 (legitimate_lo_sum_address_p): Ditto.
890 (rs6000_legitimize_address): Ditto.
891 (rs6000_legitimize_reload_address): Ditto.
892 (rs6000_legitimate_address_p): Ditto.
893 (abi_v4_pass_in_fpr): Ditto.
894 (setup_incoming_varargs): Ditto.
895 (rs6000_gimplify_va_arg): Ditto.
896 (rs6000_split_multireg_move): Ditto.
897 (rs6000_savres_strategy): Ditto.
898 (rs6000_emit_prologue_components): Ditto.
899 (rs6000_emit_epilogue_components): Ditto.
900 (rs6000_emit_prologue): Ditto.
901 (rs6000_emit_epilogue): Ditto.
902 (rs6000_elf_file_end): Ditto.
903 (rs6000_function_value): Ditto.
904 (rs6000_libcall_value): Ditto.
905 * config/rs6000/rs6000.h: Ditto.
906 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
907 (TARGET_MINMAX): ... this. New.
908 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
909 * config/rs6000/rs6000.md: Remove Xilinx FP support.
910 (*movsi_internal1_single): Delete.
911 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
912 mfpu=, mxilinx-fpu): Delete.
913 * config/rs6000/singlefp.h: Delete.
914 * config/rs6000/sysv4.h: Remove Xilinx FP support.
915 * config/rs6000/t-rs6000: Ditto.
916 * config/rs6000/t-xilinx: Delete.
917 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
918 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
919 (VStype_simple): Delete.
920 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
921 * config/rs6000/xfpu.h: Delete.
922 * config/rs6000/xfpu.md: Delete.
923 * config/rs6000/xilinx.h: Delete.
924 * config/rs6000/xilinx.opt: Delete.
925 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
926 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
928 2018-05-04 Tom de Vries <tom@codesourcery.com>
930 PR libgomp/85639
931 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
932 if ignore == 0.
934 2018-05-04 Richard Biener <rguenther@suse.de>
936 PR middle-end/85627
937 * tree-complex.c (update_complex_assignment): We are always in SSA form.
938 (expand_complex_div_wide): Likewise.
939 (expand_complex_operations_1): Likewise.
940 (expand_complex_libcall): Preserve EH info of the original stmt.
941 (tree_lower_complex): Handle removed blocks.
942 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
943 on complex multiplication and division libcall builtins.
945 2018-05-04 Richard Biener <rguenther@suse.de>
947 PR middle-end/85574
948 * fold-const.c (negate_expr_p): Restrict negation of operand
949 zero of a division to when we know that can happen without
950 overflow.
951 (fold_negate_expr_1): Likewise.
953 2018-05-04 Jakub Jelinek <jakub@redhat.com>
955 PR libstdc++/85466
956 * real.h (real_nextafter): Declare.
957 * real.c (real_nextafter): New function.
958 * fold-const-call.c (fold_const_nextafter): New function.
959 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
960 CASE_CFN_NEXTTOWARD.
961 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
962 even when arg1_mode is different from arg0_mode.
964 2018-05-03 Nathan Sidwell <nathan@acm.org>
966 * doc/extend.texi (Deprecated Features): Remove
967 -ffriend-injection.
968 (Backwards Compatibility): Likewise.
969 * doc/invoke.texi (C++ Language Options): Likewise.
970 (C++ Dialect Options): Likewise.
972 2018-05-03 Jakub Jelinek <jakub@redhat.com>
974 PR target/85530
975 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
976 _mm512_mask_mullox_epi64): New intrinsics.
978 2018-05-03 Tom de Vries <tom@codesourcery.com>
980 PR testsuite/85106
981 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
982 dump files): Add offload-tree.
984 2018-05-03 Richard Biener <rguenther@suse.de>
986 PR tree-optimization/85615
987 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
988 to loops not nested in BBs loop father to avoid creating multi-entry
989 loops.
991 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
993 PR tree-optimization/70291
994 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
995 arguments. Change return type to tree. Emit libcall as a new
996 statement rather than replacing existing one when inplace_p is true.
997 (expand_complex_multiplication_components): New function.
998 (expand_complex_multiplication): Expand floating-point complex
999 multiplication using the above.
1000 (expand_complex_division): Rename inner_type parameter to type.
1001 Update expand_complex_libcall call-site.
1002 (expand_complex_operations_1): Update expand_complex_multiplication
1003 and expand_complex_division call-sites.
1005 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1007 PR target/85582
1008 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1009 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1010 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1011 the highest significant bit of the shift count mask is clear. In
1012 check whether and[sq]i3 is needed verify that all significant bits
1013 of the shift count other than the highest are set.
1015 2018-05-02 Tom de Vries <tom@codesourcery.com>
1017 PR libgomp/82428
1018 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1019 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1020 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1021 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1022 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1023 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1024 __builtin_goacc_parlevel_size.
1026 2018-05-02 Richard Biener <rguenther@suse.de>
1028 PR tree-optimization/85597
1029 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1030 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1031 directly.
1033 2018-05-02 Tom de Vries <tom@codesourcery.com>
1035 PR testsuite/85106
1036 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1037 dump files): Add ltrans-tree.
1039 2018-05-02 Tom de Vries <tom@codesourcery.com>
1041 PR testsuite/85106
1042 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1043 dump files): Add wpa-ipa.
1045 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1047 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1048 powerpc*-*-linux*paired* target.
1049 * config/rs6000/750cl.h: Delete.
1050 * config/rs6000/paired.h: Delete.
1051 * config/rs6000/paired.md: Delete.
1052 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1053 float support.
1054 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1055 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1056 comment. Remove paired float support.
1057 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1058 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1059 VECTOR_PAIRED.
1060 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1061 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1062 declarations.
1063 * config/rs6000/rs6000.c: Remove paired float support.
1064 (paired_expand_vector_init, paired_expand_vector_move,
1065 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1066 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1067 paired_expand_builtin, paired_expand_predicate_builtin,
1068 paired_init_builtins): Delete.
1069 * config/rs6000/rs6000.h: Remove paired float support.
1070 * config/rs6000/rs6000.md: Remove paired float support.
1071 (move_from_CR_ov_bit): Delete.
1072 * config/rs6000/rs6000.opt (mpaired): Delete.
1073 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1074 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1076 2018-05-02 Richard Biener <rguenther@suse.de>
1078 PR middle-end/85567
1079 * gimplify.c (gimplify_save_expr): When in SSA form allow
1080 SAVE_EXPRs to compute to SSA vars.
1082 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1084 PR target/85582
1085 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1086 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1087 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1088 clobber operands[2], instead use a new pseudo. Formatting fixes.
1090 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1092 PR tree-optimization/85586
1093 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1094 exit early for statements in the same group if the accesses are
1095 not strided.
1097 2018-05-02 Tom de Vries <tom@codesourcery.com>
1099 PR lto/85451
1100 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1101 error message.
1103 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1105 PR tree-optimization/85143
1106 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1108 2018-05-01 Tom de Vries <tom@codesourcery.com>
1110 PR lto/85451
1111 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1112 not found" error message.
1114 2018-05-01 Tom de Vries <tom@codesourcery.com>
1116 PR other/83786
1117 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1118 * vec.c (test_ordered_remove_if): New function.
1119 (vec_c_tests): Call test_ordered_remove_if.
1120 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1121 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1122 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1123 VEC_ORDERED_REMOVE_IF.
1125 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1127 PR tree-optimization/82665
1128 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1129 pointer subtraction where arguments come from a memchr call.
1131 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1133 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1134 --push-state --as-needed and --pop-state instead of --as-needed and
1135 --no-as-needed if ld supports it.
1136 * configure: Regenerated.
1138 PR web/85578
1139 * doc/install.texi2html: Replace _002d with - and _002a with * in
1140 generated html files using sed.
1142 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1144 PR c++/85523
1145 * gcc-rich-location.c (blank_line_before_p): New function.
1146 (use_new_line): New function.
1147 (gcc_rich_location::add_fixit_insert_formatted): New function.
1148 * gcc-rich-location.h
1149 (gcc_rich_location::add_fixit_insert_formatted): New function.
1151 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1153 * selftest.c (assert_streq): Rename "expected" and "actual" to
1154 "val1" and "val2". Extend NULL-handling to cover both inputs
1155 symmetrically, while still requiring both to be non-NULL for a pass.
1156 * selftest.h (assert_streq): Rename "expected" and "actual" to
1157 "val1" and "val2".
1158 (ASSERT_EQ): Likewise.
1159 (ASSERT_EQ_AT): Likewise.
1160 (ASSERT_KNOWN_EQ): Likewise.
1161 (ASSERT_KNOWN_EQ_AT): Likewise.
1162 (ASSERT_NE): Likewise.
1163 (ASSERT_MAYBE_NE): Likewise.
1164 (ASSERT_MAYBE_NE_AT): Likewise.
1165 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1166 the assertion to pass.
1167 (ASSERT_STREQ_AT): Likewise.
1169 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1171 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1172 interaction with -pie.
1174 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1176 * selftest.h: Fix alphabetization of per-source-file selftest
1177 declarations.
1179 2018-04-30 Jason Merrill <jason@redhat.com>
1181 PR c++/61982 - dead stores to destroyed objects.
1182 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1183 of clobber.
1185 2018-04-30 Jason Merrill <jason@redhat.com>
1187 * tree.c (build_clobber): New.
1188 * tree.h: Declare it.
1189 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1191 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1193 * diagnostic-show-locus.c (layout::layout): Update for
1194 location_get_source_line returning a char_span.
1195 (struct char_span): Move to input.h.
1196 (struct correction): Update for fields in char_span becoming
1197 private.
1198 (struct source_line): Update for location_get_source_line
1199 returning a char_span.
1200 (layout::print_line): Likewise.
1201 * edit-context.c (edited_file::print_content): Likewise.
1202 (edited_file::print_diff_hunk): Likewise.
1203 (edited_file::print_run_of_changed_lines): Likewise.
1204 (edited_file::get_num_lines): Likewise.
1205 (edited_line::edited_line): Likewise.
1206 * final.c (asm_show_source): Likewise.
1207 * input.c (location_get_source_line): Convert return type
1208 from const char * to char_span, losing the final "line_len"
1209 param.
1210 (dump_location_info): Update for the above.
1211 (get_substring_ranges_for_loc): Likewise. Use a char_span
1212 when handling the literal within the line.
1213 (test_reading_source_line): Update for location_get_source_line
1214 returning a char_span.
1215 * input.h (class char_span): Move here from
1216 diagnostic-show-locus.c, converting from a struct to a class.
1217 Make data members private.
1218 (char_span::operator bool): New.
1219 (char_span::length): New.
1220 (char_span::get_buffer): New.
1221 (char_span::operator[]): New.
1222 (char_span::subspan): Make const.
1223 (char_span::xstrdup): New.
1224 (location_get_source_line): Convert return type from const char *
1225 to char_span, losing the final "line_size" param.
1227 2018-04-30 Jan Hubicka <jh@suse.cz>
1229 * lto-wrapper.c (ltrans_priorities): New static var.
1230 (cmp_priority): New.
1231 (run_gcc): Read priorities and if doing parallel build order
1232 the Makefile by them.
1234 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1236 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1238 2018-04-30 Richard Biener <rguenther@suse.de>
1240 * tree-cfg.c (verify_address): Remove base argument, add
1241 flag whether to check TREE_ADDRESSABLE and do that.
1242 (verify_expr): Remove.
1243 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1244 (verify_gimple_assign_single): Likewise.
1245 (verify_gimple_switch): Likewise.
1246 (verify_expr_location_1): Dereference tp once. Add (disabled)
1247 piece from verify_expr.
1248 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1250 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1252 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1254 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1256 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1257 (small_data_pattern): Likewise.
1258 (arc_rewrite_small_data): Likewise.
1259 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1260 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1261 (get_symbol_alignment): New function.
1262 (legitimate_small_data_address_p): Likewise.
1263 (legitimate_scaled_address): Update, call
1264 legitimate_small_data_address_p.
1265 (output_sdata): New static variable.
1266 (arc_print_operand): Update how we handle small data operands.
1267 (arc_print_operand_address): Likewise.
1268 (arc_legitimate_address_p): Update, use
1269 legitimate_small_data_address_p.
1270 (arc_rewrite_small_data_p): Remove.
1271 (arc_rewrite_small_data_1): Likewise.
1272 (arc_rewrite_small_data): Likewise.
1273 (small_data_pattern): Likewise.
1274 (compact_sda_memory_operand): Update to use
1275 legitimate_small_data_address_p and get_symbol_alignment.
1276 (prepare_move_operands): Don't rewite sdata pattern.
1277 (prepare_extend_operands): Remove.
1278 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1279 pattern.
1280 (zero_extendqisi2): Likewise.
1281 (zero_extendhisi2): Likewise.
1282 (extendqihi2): Likewise.
1283 (extendqisi2): Likewise.
1284 (extendhisi2): Likewise.
1285 (addsi3): Likewise.
1286 (subsi3): Likewise.
1287 (andsi3): Likewise.
1288 * config/arc/constraints.md (Usd): Change it to memory constraint.
1290 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1292 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1293 as source of std instructions.
1294 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1295 allow 6-bit constants as source for store instructions.
1296 (movdi_insn): Update instruction pattern to allow 6-bit constants
1297 as source for store instructions.
1299 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1301 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1303 2018-04-30 Nathan Sidwell <nathan@acm.org>
1304 Sandra Loosemore <sandra@codesourcery.com>
1306 * dumpfile.c (dump_open): Allow '-' for stdout.
1307 * doc/invoke.texi (Developer Options): Document dump filename
1308 determination early. Document stdin/stdout selection.
1310 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1312 Microblaze Target: PIC data text relative
1314 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1315 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1316 Add declaration.
1317 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1318 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1319 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1320 New addressing mode for data-text relative position indepenedent code.
1321 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1322 'ADDRESS_SYMBOLIC_TXT_REL'.
1323 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1324 (microblaze_legitimate_pic_operand): Exclude function calls from
1325 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1326 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1327 addresses cases.
1328 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1329 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1330 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1331 for 'address + offset'.
1332 (microblaze_expand_prologue): Add new function prologue call for
1333 'r20' assignation.
1334 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1335 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1336 table in case of TARGET_PIC_DATA_TEXT_REL.
1337 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1338 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1339 Add new macros 'UNSPEC_TEXT',
1340 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1341 + exclude function calls from 'UNSPEC_PLT' in case of data text
1342 relative mode.
1343 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1344 new target hook for generating address diff vector tables in case of
1345 flag_pic.
1346 * doc/tm.texi : Regenerate.
1347 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1348 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1349 of addr diff vector generation.
1350 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1351 target hook definition.
1352 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1353 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1354 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1355 option for data text relative.
1357 2018-04-30 Richard Biener <rguenther@suse.de>
1359 * tree-chrec.h (evolution_function_is_constant_p): Remove
1360 redundant check.
1361 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1363 2018-04-30 Richard Biener <rguenther@suse.de>
1365 PR bootstrap/85571
1366 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1368 2018-04-30 Richard Biener <rguenther@suse.de>
1370 PR tree-optimization/28364
1371 PR tree-optimization/85275
1372 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1373 copying first exit test.
1375 2018-04-28 Mark Wielaard <mark@klomp.org>
1377 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1378 dwarf_version >= 5.
1379 (dwarf_AT): Handle DW_AT_addr_base.
1380 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1382 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1384 PR target/84431
1385 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1386 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1387 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1388 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1390 2018-04-28 Richard Biener <rguenther@suse.de>
1392 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1393 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1394 to reflect use. Only add interesting stmts.
1396 2018-04-27 Martin Jambor <mjambor@suse.cz>
1398 PR ipa/85549
1399 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1400 the jump function allows for passing through aggregate values.
1402 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1404 * input.h (in_system_header_at): Convert from macro to inline
1405 function.
1406 (from_macro_expansion_at): Likewise.
1407 (from_macro_definition_at): Likewise.
1409 2018-04-27 Jeff Law <law@redhat.com>
1411 * config.gcc: Mark tile* targets as deprecated/obsolete.
1413 2018-04-27 Richard Biener <rguenther@suse.de>
1415 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1416 fix for ILP32.
1418 2018-04-27 Richard Biener <rguenther@suse.de>
1420 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1422 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1424 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1425 with Yd constraint. Set "preferred_for_speed" attribute from
1426 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1427 with Yd constraint.
1428 (*movdi_internal): Ditto.
1429 (movti_interunit splitters): Remove
1430 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1431 (movdi_interunit splitters): Ditto.
1432 * config/i386/constraints.md (Ye): Remove.
1433 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1435 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1437 PR target/85512
1438 * config/aarch64/constraints.md (Usg): Limit to 31.
1439 (Usj): Limit to 63.
1441 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1443 PR tree-optimization/85529
1444 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1445 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1446 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1447 zero extension or masking of the MSB bit.
1448 (optimize_range_tests): Add FIRST_BB argument, pass it through
1449 to optimize_range_tests_var_bound.
1450 (maybe_optimize_range_tests, reassociate_bb): Adjust
1451 optimize_range_tests callers.
1453 2018-04-26 Richard Biener <rguenther@suse.de>
1454 Jakub Jelinek <jakub@redhat.com>
1456 * cgraph.h (symbol_table): Just declare debug method here.
1457 * symtab.c (symbol_table::debug): Define.
1459 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1461 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1463 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1465 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1466 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1467 (*movdi_internal): Substitute Yi and Yj constraint with x
1468 and Ym and Yn constraint with y constraint. Update "isa"
1469 attribute and set "preferred_for_speed" attribute from
1470 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1471 (*movsi_internal): Ditto.
1472 (*movdf_internal): Ditto.
1473 (*movsf_internal): Ditto.
1474 (*zero_extendsidi2): Ditto.
1475 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1476 (sse2_loadld): Ditto.
1477 (*vec_extract<ssevecmodelower>_0): Ditto.
1478 (*vec_extractv4si_0_zext_sse4): Ditto.
1479 (vec_concatv2di): Ditto.
1480 (*vec_dup<mode>): Ditto.
1481 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1482 * config/i386/constraints.md (Yi): Remove.
1483 (Yj): Remove.
1484 (Ym): Remove.
1485 (Yn): Remove.
1487 2018-04-26 Nathan Sidwell <nathan@acm.org>
1489 * dumpfile.c (dump_open): New.
1490 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1491 (dump_finish): Detect stdio/stderr by value not name.
1493 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1495 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1497 2018-04-26 Tom de Vries <tom@codesourcery.com>
1499 PR target/84952
1500 * config/nvptx/nvptx.c (verify_neutering_jumps)
1501 (verify_neutering_labels): New function
1502 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1504 2018-04-26 Tom de Vries <tom@codesourcery.com>
1506 PR target/84025
1507 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1508 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1509 need neutering.
1511 2018-04-26 Richard Biener <rguenther@suse.de>
1512 Tom de Vries <tom@codesourcery.com>
1514 PR lto/85422
1515 * lto-streamer-out.c (output_function): Fixup loops if required to match
1516 discovery done in the reader.
1518 2018-04-26 Richard Biener <rguenther@suse.de>
1520 PR tree-optimization/85116
1521 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1522 have a loop exit from the single latch predecessor. Remove
1523 case of header with just condition.
1524 (ch_base::copy_headers): Exclude infinite loops from any
1525 processing.
1526 (pass_ch::execute): Record exits.
1528 2018-04-26 Richard Biener <rguenther@suse.de>
1530 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1531 prologue cost vector and pass it to vect_get_load_cost.
1532 (vect_get_peeling_costs_all_drs): Likewise.
1533 (vect_peeling_hash_get_lowest_cost): Likewise.
1534 (vect_enhance_data_refs_alignment): Likewise.
1536 2018-04-26 Richard Biener <rguenther@suse.de>
1538 PR middle-end/85450
1539 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1540 checking of integer<->pointer conversions.
1541 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1542 sign-/zero-extending pointer types.
1543 (expand_omp_for_static_chunk): Likewise.
1545 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1546 Jean Lee <xiaoyur347@gmail.com>
1548 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1549 (TARGET_ASAN_SHADOW_OFFSET): Define.
1550 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1551 true for -fsanitize=address.
1553 2018-04-25 Mark Wielaard <mark@klomp.org>
1555 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1556 shorter ones.
1558 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1560 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1561 than "alu", remove explicit "memory" and "imm_disp" attributes.
1562 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1564 PR middle-end/85414
1565 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1566 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1567 gen_lowpart_no_emit.
1569 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1571 PR target/85473
1572 * config/i386/i386.c (ix86_expand_builtin): Change memory
1573 operand to XI, extend p0 to Pmode.
1574 * config/i386/i386.md: Change unspec volatile and operand
1575 1 mode to XI, change operand 0 mode to P.
1577 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1579 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1580 GET_MODE_MASK before any checking.
1581 (nds32_can_use_bset_p): Likewise.
1582 (nds32_can_use_btgl_p): Likewise.
1584 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1586 * config/nds32/nds32-doubleword.md: New define_split pattern for
1587 illegal register number.
1589 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1591 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1593 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1595 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1597 2018-04-25 Richard Biener <rguenther@suse.de>
1599 * lto-streamer.h (LTO_major_version): Bump to 8.
1601 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1603 * BASE-VER: Set to 9.0.0.
1605 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1607 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1608 in __abskf2 and __powikf2.
1610 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1612 PR target/85512
1613 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1614 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1615 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1616 Use the above on operand 2. Reindent.
1617 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1619 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1621 PR target/85485
1622 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1623 handle OPT_mcet.
1624 * config/i386/i386.opt (mcet): Removed.
1625 * doc/install.texi: Remove -mcet documentation.
1626 * doc/invoke.texi: Likewise.
1628 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1630 PR target/85485
1631 * doc/install.texi: Remove -mcet from bootstrap-cet.
1633 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1635 PR target/85511
1636 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1637 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1638 if TARGET_64BIT.
1640 PR target/85503
1641 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1642 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1643 containing a CONST_VECTOR.
1645 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1647 * doc/install.texi: Update newlib dependency for nvptx.
1649 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1651 PR target/85508
1652 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1653 instead of INTVAL when shifting x left.
1655 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1657 PR tree-optimization/85478
1658 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1659 vect_grouped_store_supported for single element vectors.
1661 2018-04-24 Richard Biener <rguenther@suse.de>
1663 PR target/85491
1664 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1665 load cost increase to the case of non-constant step.
1667 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1669 PR target/84828
1670 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1671 destination if any_malformed_asm.
1673 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1675 PR middle-end/85496
1676 * expr.c (store_field): In the bitfield case, if the value comes from
1677 a function call and is returned in registers by means of a PARALLEL,
1678 do not change the mode of the temporary unless BLKmode and VOIDmode.
1680 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1682 PR rtl-optimization/85423
1683 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1684 dependencies to debug insns when the previous insn is non-debug.
1686 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1688 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1689 enums into a single definition.
1690 (fls): Fix predicates and printing.
1691 (seti): Likewise.
1693 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1695 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1696 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1697 and short u6 immediate.
1698 (check_if_valid_sleep_operand): Remove.
1699 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1701 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1703 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1704 flag_always_save_lp condition.
1705 * config/nds32/nds32.opt (malways-save-lp): New option.
1707 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1709 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1710 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1711 * config/nds32/nds32.h
1712 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1713 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1715 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1717 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1718 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1720 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1721 Chung-Ju Wu <jasonwucj@gmail.com>
1723 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1724 Declare.
1725 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1726 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1728 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1730 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1732 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1734 * config/nds32/nds32-protos.h (nds32_data_alignment,
1735 nds32_local_alignment): Declare.
1736 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1737 nds32_local_alignment): New functions.
1738 (TARGET_CONSTANT_ALIGNMENT): Define.
1739 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1741 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1743 * config/nds32/nds32.c
1744 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1745 (TARGET_MODES_TIEABLE_P): Likewise.
1747 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1749 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1750 level Ofast and Og.
1752 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1753 Chung-Ju Wu <jasonwucj@gmail.com>
1755 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1756 for unaligned access.
1757 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1758 unaligned access.
1759 * config/nds32/nds32-intrinsic.md: Likewise.
1760 * config/nds32/nds32_intrinsic.h: Likewise.
1761 * config/nds32/nds32.h (nds32_builtins): Likewise.
1762 * config/nds32/nds32.opt (munaligned-access): New option.
1763 * config/nds32/nds32.c (nds32_asm_file_start): Display
1764 flag_unaligned_access status.
1766 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1768 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1769 -mno-relax is present.
1770 * config/riscv/linux.h (LINK_SPEC): Ditto.
1772 2018-04-20 Martin Sebor <msebor@redhat.com>
1774 PR c/85365
1775 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1776 for null pointers.
1777 (gimple_fold_builtin_stxcpy_chk): Same.
1778 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1780 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1782 PR target/85456
1783 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1784 __powikf2 when long double is IEEE 128-bit.
1786 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1788 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1789 step to make sure stack always aligned.
1791 2018-04-20 Carl Love <cel@us.ibm.com>
1793 PR target/83402
1794 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1795 size check for arg0.
1797 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1798 Tom de Vries <tom@codesourcery.com>
1800 PR target/85445
1801 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1802 Emit insns for calls too.
1803 (nvptx_find_par): Always look for worker-level predecessor insn.
1804 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1805 calls.
1806 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1807 (nvptx_process_pars): Propagate frames for calls.
1809 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1811 PR target/85469
1812 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1813 Removed.
1814 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1815 (ix86_handle_option): Don't handle OPT_mibt.
1816 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1817 __SHSTK__.
1818 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1819 has_ibt and ibt.
1820 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1821 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1822 (ix86_target_macros): Define __CET__ with flag_cf_protection
1823 for -fcf-protection.
1824 * config/i386/i386.c (isa2_opts): Remove -mibt.
1825 * config/i386/i386.h (TARGET_IBT): Removed.
1826 (TARGET_IBT_P): Likewise.
1827 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1828 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1829 * config/i386/i386.opt (mcet): Update help message.
1830 (mshstk): Likewise.
1831 (mibt): Removed.
1832 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1833 -mcet as an alias for -mshstk.
1835 2018-04-20 Richard Biener <rguenther@suse.de>
1837 PR middle-end/85475
1838 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1839 complexity by forcing a single use of the multiply operand.
1841 2018-04-20 Martin Jambor <mjambor@suse.cz>
1843 ipa/85449
1844 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1845 recursion dependency to only apply to non-clones.
1847 2018-04-20 Martin Jambor <mjambor@suse.cz>
1849 ipa/85447
1850 * ipa-cp.c (create_specialized_node): Check that clones of
1851 self-recursive edges exist during IPA-CP.
1853 2018-04-19 Toon Moene <toon@moene.org>
1855 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1856 by -O3.
1858 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1860 PR tree-optimization/85467
1861 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1862 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1863 VECTOR_CST element to type.
1865 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1867 PR target/85397
1868 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1869 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1870 (builtin_longjmp): Likewise.
1871 (save_stack_nonlocal): New pattern.
1872 (restore_stack_nonlocal): Likewise.
1874 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1876 PR target/85404
1877 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1878 Replace ASM_OUTPUT_LABEL with fprintf.
1880 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1882 PR target/85417
1883 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1884 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1885 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1886 define __IBT__ and __SHSTK__ for -fcf-protection.
1887 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1888 TARGET_IBT.
1889 (ix86_trampoline_init): Likewise.
1890 (x86_output_mi_thunk): Likewise.
1891 (ix86_notrack_prefixed_insn_p): Likewise.
1892 (ix86_option_override_internal): Don't disallow -fcf-protection.
1893 * config/i386/i386.md (rdssp<mode>): Also enable for
1894 -fcf-protection.
1895 (incssp<mode>): Likewise.
1896 (nop_endbr): Likewise.
1897 * config/i386/i386.opt (mcet): Change help message to built-in
1898 functions only.
1899 (mibt): Likewise.
1900 (mshstk): Likewise.
1901 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1902 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1903 enable CET built-in functions.
1905 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1907 * common/config/i386/i386-common.c
1908 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1909 OPTION_MASK_ISA_MOVDIRI_UNSET,
1910 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1911 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1912 * config.gcc (movdirintrin.h): New header.
1913 * config/i386/cpuid.h (bit_MOVDIRI,
1914 bit_MOVDIR64B): New bits.
1915 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1916 and -mmvodir64b.
1917 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1918 (VOID, PVOID, PCVOID)): New function types.
1919 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1920 __builtin_ia32_directstoreu_u64,
1921 __builtin_ia32_movdir64b): New builtins.
1922 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1923 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1924 and -mmovdiri.
1925 (ix86_valid_target_attribute_inner_p): Ditto.
1926 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1927 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1928 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1929 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1930 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1931 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1932 (movdiri<mode>, movdir64b_<mode>): New.
1933 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1934 * config/i386/immintrin.h: Include movdirintrin.h.
1935 * config/i386/movdirintrin.h: New file.
1936 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1938 2018-04-19 Richard Biener <rguenther@suse.de>
1940 PR middle-end/85455
1941 * cfg.c (clear_bb_flags): When loop state says we have
1942 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1944 2018-04-19 Richard Biener <rguenther@suse.de>
1946 PR tree-optimization/84737
1947 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1948 copying restrict info.
1949 (vect_setup_realignment): Use it.
1950 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1951 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1952 the first DR to all generated stores.
1953 (vectorizable_load): Likewise for loads.
1955 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1957 PR tree-optimization/85446
1958 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1959 the integral and pointer types to have the same precision.
1961 * doc/install.texi: Document --disable-cet being the default and
1962 --enable-cet=auto.
1964 2018-04-18 Martin Liska <mliska@suse.cz>
1966 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1967 style.
1969 2018-04-18 Martin Liska <mliska@suse.cz>
1971 Revert
1972 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1974 PR ipa/83983
1975 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1976 arguments if they are comparable.
1978 2018-04-18 Martin Liska <mliska@suse.cz>
1980 Revert
1981 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1983 PR lto/84805
1984 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1985 incomplete types.
1987 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1989 PR target/85388
1990 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1991 ENDBR after calling __morestack.
1993 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1995 PR jit/85384
1996 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1997 by using gcc_base_ver to generate a gcc_driver_version, and use
1998 it when generating GCC_DRIVER_NAME.
1999 * configure: Regenerate.
2001 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2003 PR target/81084
2004 * config.gcc: Obsolete powerpc*-*-*spe*.
2006 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2008 PR debug/84637
2009 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2010 (stabstr_D): Change type of unum from unsigned int to
2011 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2012 type.
2014 2018-04-17 Jim Wilson <jimw@sifive.com>
2016 PR 84856
2017 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2018 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2019 Set arg_pointer_offset after using pretend_args_size.
2021 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2023 PR rtl-optimization/85431
2024 * dse.c (record_store): Ignore zero width stores.
2026 PR sanitizer/85230
2027 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2028 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2029 __builtin_stack_restore rather than after it.
2030 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2031 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2032 argument instead of virtual_dynamic_stack_rtx.
2034 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2036 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2037 New prototype.
2038 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2039 Add note to error message to explain internal mapping of overloaded
2040 built-in function name to non-overloaded built-in function name.
2041 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2042 function.
2044 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2046 PR target/85424
2047 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2048 where the inputs overlap with the output.
2050 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2052 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2053 (=v, v) alternative and explicit "memory" attribute.
2054 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2055 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2056 attributes.
2057 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2058 "sselog1" type instead of "sselog".
2059 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2060 "sselog". Remove explicit "memory" attribute.
2061 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2062 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2063 attributes.
2064 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2065 "sselog1" type instead of "sselog". Remove explicit "memory"
2066 attribute.
2067 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2068 use "sselog1" type instead of "sselog". Remove explicit "memory"
2069 attribute.
2070 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2071 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2072 attributes.
2073 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2074 "sselog1" type instead of "sselog". Remove explicit "memory"
2075 attribute.
2076 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2077 use "sselog1" type instead of "sselog". Remove explicit "memory"
2078 attribute.
2080 PR target/85430
2081 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2083 PR middle-end/85414
2084 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2085 on a SUBREG.
2087 2018-04-17 Martin Jambor <mjambor@suse.cz>
2089 PR ipa/85421
2090 * ipa-cp.c (create_specialized_node): Call
2091 expand_all_artificial_thunks if necessary.
2093 2018-04-17 Martin Liska <mliska@suse.cz>
2095 PR lto/85405
2096 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2097 in message, remote space in between '_G' and '('.
2099 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2101 PR target/85281
2102 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2103 avx512f_vmcmp<mode>3<round_saeonly_name>,
2104 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2105 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2106 avx512f_rndscale<mode><round_saeonly_name>,
2107 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2108 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2109 Use %<iptr>2 instead of %2 for -masm=intel.
2110 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2111 avx512f_vcvttss2usi<round_saeonly_name>,
2112 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2113 -masm=intel.
2114 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2115 avx512f_vcvttsd2usi<round_saeonly_name>,
2116 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2117 Use %q1 instead of %1 for -masm=intel.
2118 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2119 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2120 of %3 for -masm=intel.
2121 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2122 -masm=intel.
2123 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2124 -masm=intel.
2125 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2126 -masm=intel.
2127 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2128 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2129 %g1.
2130 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2131 -masm=intel.
2132 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2133 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2134 %g1 and one with %0 and %1.
2135 (avx512er_vmrcp28<mode><round_saeonly_name>,
2136 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2137 %1 for -masm=intel.
2138 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2139 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2140 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2141 of %0 and %{%4%} for -masm=intel.
2142 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2143 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2144 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2145 order of %0 and %{%5%}%{z%} for -masm=intel.
2147 2018-04-17 Jan Hubicka <jh@suse.cz>
2149 PR lto/85405
2150 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2152 2018-04-17 Martin Liska <mliska@suse.cz>
2154 PR ipa/85329
2155 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2156 for target_clone error message. Make default implementation
2157 clone to be a local declaration.
2158 (separate_attrs): Add new argument and check for an empty
2159 string.
2160 (expand_target_clones): Handle it.
2161 (ipa_target_clone): Make redirection just for target_clones
2162 functions.
2164 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2165 Tom de Vries <tom@codesourcery.com>
2167 PR middle-end/84955
2168 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2169 tiled basic blocks without omp continue statements.
2171 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2173 PR target/83660
2174 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2175 vec_extract expression as having side effects to make sure it gets
2176 a cleanup point.
2178 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2180 PR target/85403
2181 * config/i386/i386.c (get_builtin_code_for_version): Check
2182 error_mark_node.
2184 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2186 PR target/84331
2187 * gcc/config.gcc: Support "skylake".
2188 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2189 PROCESSOR_SKYLAKE.
2190 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2191 (processor_target_table): Add "skylake".
2192 (ix86_option_override_internal): Add "skylake".
2193 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2194 PROCESSOR_CANNONLAKE.
2195 (get_builtin_code_for_version): Fix priority for
2196 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2197 PROCESSOR_SKYLAKE-AVX512.
2198 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2199 (processor_type): Add PROCESSOR_SKYLAKE.
2201 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2202 Jason Merrill <jason@redhat.com>
2204 PR c++/85112
2205 * convert.c (convert_to_integer_1): Use direct recursion for
2206 enumeral types and types with a precision less than the number
2207 of bits in their mode.
2209 2018-04-16 Julia Koval <julia.koval@intel.com>
2211 PR target/84413
2212 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2213 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2215 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2217 PR target/85293
2218 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2219 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2220 and -mno-direct-move.
2222 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2224 PR target/83402
2225 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2226 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2227 Ensure negative shifts result in {0}.
2229 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2231 PR rtl-optimization/79916
2232 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2233 regs (if any) to define how to gnerate SD moves when LRA is in
2234 progress.
2236 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2238 PR rtl-optimization/85393
2239 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2240 * except.c (expand_dw2_landing_pad_for_region): Make static.
2241 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2242 a label and unconditional jump to old_bb, rather than
2243 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2244 basic block.
2246 PR rtl-optimization/85376
2247 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2248 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2249 instead of a specific value.
2251 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2252 Bin Cheng <bin.cheng@arm.com>
2254 PR tree-optimization/82965
2255 PR tree-optimization/83991
2256 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2257 by_profile_only parameter.
2258 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2259 information if the loop was predicted to iterate too many times.
2260 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2262 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2264 PR lto/71991
2265 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2266 always inline.
2268 2018-04-13 Martin Liska <mliska@suse.cz>
2269 Jakub Jelinek <jakub@redhat.com>
2271 PR middle-end/81657
2272 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2273 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2274 * builtins.c (expand_builtin_memory_copy_args): Use
2275 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2276 handle dest_addr == pc_rtx.
2278 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2280 PR target/85291
2281 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2282 asked to not generate direct moves.
2283 (fix_trunc<mode>si2_stfiwx): Similar.
2284 (fix_trunc<mode>si2_internal): Similar.
2286 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2288 PR debug/83157
2289 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2290 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2291 lookup if dest in some wider mode is known to be const0_rtx and
2292 if so, record permanent equivalence for it to be ZERO_EXTEND of
2293 the narrower mode destination.
2295 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2297 * lto-streamer-out.c (output_function): Revert 259346.
2298 * omp-expand.c (expand_oacc_for): Likewise.
2300 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2302 PR rtl-optimization/85354
2303 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2304 * sel-sched.c (sel_global_init): ... here.
2306 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2308 PR target/85238
2309 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2310 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2311 mode for PE-COFF targets.
2312 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2313 (i386_pe_asm_lto_end): Likewise.
2314 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2315 (TARGET_ASM_LTO_END): Likewise.
2316 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2317 (i386_pe_asm_lto_start): New function.
2318 (i386_pe_asm_lto_end): Likewise.
2320 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2321 Richard Biener <rguenther@suse.de>
2323 PR middle-end/84955
2324 * lto-streamer-out.c (output_function): Fix CFG loop state before
2325 streaming out.
2326 * omp-expand.c (expand_oacc_for): Handle calls to internal
2327 functions like regular functions.
2329 2018-04-12 Richard Biener <rguenther@suse.de>
2331 PR lto/85371
2332 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2333 for the early LTO debug to properly generate references to it
2334 during DIE emission. Do not re-use that for the skeleton for
2335 split-dwarf.
2336 (dwarf2out_early_finish): Likewise.
2338 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2340 PR target/85328
2341 * config/i386/sse.md
2342 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2343 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2344 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2345 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2346 and output is a reg, avoid creating invalid lowpart subreg, but
2347 instead split into a 512-bit move. Don't split if not AVX512VL,
2348 input is xmm16+ reg and output is a mem.
2349 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2350 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2351 xmm16+ reg and output is a mem.
2353 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2355 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2356 also for flag_dwarf2_cfi_asm.
2358 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2360 PR rtl-optimization/85342
2361 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2362 a bool scalar var inside of the loop instead. Don't try to update
2363 recog_data.operand after failed apply_change_group.
2365 2018-04-12 Tom de Vries <tom@codesourcery.com>
2367 PR target/85296
2368 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2369 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2370 array with flexible array member as array without given dimension.
2371 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2372 argument for undefined param to true.
2374 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2376 PR target/85321
2377 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2378 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2379 from PowerPC section.
2380 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2381 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2382 help text that is too long.
2383 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2384 help text that is too long.
2385 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2386 help text that is too long.
2388 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2390 * config/alpha/alpha.md (stack_probe_internal): Rename
2391 from "probe_stack". Update all callers.
2393 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2395 PR rtl-optimization/84566
2396 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2397 sched_macro_fuse_insns.
2399 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2401 PR target/84301
2402 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2403 (compute_block_dependences): ... from here.
2405 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2407 PR tree-optimization/85331
2408 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2409 from int to HOST_WIDE_INT.
2411 2018-04-11 Martin Jambor <mjambor@suse.cz>
2413 PR ipa/84149
2414 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2415 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2416 not the same as the source val.
2417 (cgraph_edge_brings_value_p): New parameter.
2418 (gather_edges_for_value): Pass destination value to
2419 cgraph_edge_brings_value_p.
2420 (perhaps_add_new_callers): Likewise.
2421 (get_info_about_necessary_edges): Likewise and exclude values brought
2422 only by self-recursive edges.
2423 (create_specialized_node): Redirect only clones of self-calling edges.
2424 (+self_recursive_pass_through_p): New function.
2425 (find_more_scalar_values_for_callers_subset): Use it.
2426 (find_aggregate_values_for_callers_subset): Likewise.
2427 (known_aggs_to_agg_replacement_list): Removed.
2428 (decide_whether_version_node): Re-calculate known constants for all
2429 remaining context clones.
2431 2018-04-11 Richard Biener <rguenther@suse.de>
2433 PR lto/85339
2434 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2435 from early DWARF output.
2436 (dwarf2out_early_finish): Output line info unconditionally into
2437 early DWARF and add reference to it.
2439 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2441 PR target/85281
2442 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2443 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2444 other than V2DFmode using iptr mode attribute.
2445 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2447 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2449 PR rtl-optimization/84659
2450 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2452 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2454 PR debug/85302
2455 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2456 SIZEP is NULL.
2457 (output_loc_list): Pass address of a dummy size variable even in the
2458 locview handling loop.
2459 (index_location_lists): Add comment on why skip_loc_list_entry can't
2460 call size_of_locs.
2462 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2464 PR target/85261
2465 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2466 into register.
2468 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2470 PR target/85321
2471 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2472 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2473 and -mstring-compare-inline-limit.
2475 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2477 PR target/85287
2478 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2479 for stack clash protection in a register whenever we need it to be in
2480 a register.
2482 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2484 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2485 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2487 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2489 PR target/85321
2490 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2491 the help text.
2492 (mlong-double-): Ditto.
2493 * config/rs6000/sysv4.opt (msdata=): Ditto.
2494 (mtls-size=): Ditto.
2496 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2498 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2499 erroneous entries for
2500 "vector int vec_ldl (int, long int *)", and
2501 "vector unsigned int vec_ldl (int, unsigned long int *)".
2502 Add comments and entries for
2503 "vector bool char vec_ldl (int, bool char *)",
2504 "vector bool short vec_ldl (int, bool short *)",
2505 "vector bool int vec_ldl (int, bool int *)",
2506 "vector bool long long vec_ldl (int, bool long long *)",
2507 "vector pixel vec_ldl (int, pixel *)",
2508 "vector long long vec_ldl (int, long long *)",
2509 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2510 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2511 type tree bool_long_long_type_node and correct definition of
2512 bool_V2DI_type_node to make reference to this new type tree.
2513 (rs6000_mangle_type): Replace erroneous reference to
2514 bool_long_type_node with bool_long_long_type_node.
2515 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2516 comments to emphasize sign distinctions for char and int types and
2517 replace RS6000_BTI_bool_long constant with
2518 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2519 use of RS6000_BTI_pixel.
2520 (bool_long_type_node): Remove this macro definition.
2521 (bool_long_long_type_node): New macro definition
2523 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2525 PR rtl-optimization/85300
2526 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2527 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2528 simplify_unary_operation fails.
2530 2018-04-10 Martin Liska <mliska@suse.cz>
2532 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2533 cgraph_edge and ipa_ref.
2535 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2537 PR target/85177
2538 PR target/85255
2539 * config/i386/sse.md
2540 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2541 computation of the VEC_MERGE selector from mask.
2542 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2543 Fix decoding of the VEC_MERGE selector into mask.
2545 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2547 PR tree-optimization/85286
2548 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2550 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2552 * final.c (final_1): Set insn_last_address as well as
2553 insn_current_address.
2555 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2557 PR target/85173
2558 * explow.c (emit_stack_probe): Call validize_mem on memory location
2559 before passing it to gen_probe_stack. Create address operand and
2560 legitimize it for the probe_stack_address case.
2562 2018-04-09 Jan Hubicka <jh@suse.cz>
2564 PR lto/85078
2565 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2566 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2567 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2568 walk basetypes.
2569 (free_lang_data): Rebuild type inheritance graph.
2571 2018-04-09 Martin Sebor <msebor@redhat.com>
2573 * invoke.texi (-finline-small-functions): Mention other optimization
2574 options.
2575 (-findirect-inlining, -fpartial-inlining): Same.
2576 (-finline-functions-called-once): Same.
2577 (-freorder-blocks-and-partition): Same.
2579 2018-04-09 Jan Hubicka <jh@suse.cz>
2581 PR rtl/84058
2582 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2583 jumps; choose last target that matches the criteria (i.e.
2584 no partition changes for non-crossing jumps).
2585 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2586 support for redirecting crossing jumps to non-crossing.
2588 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2590 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2591 also for naked functions.
2593 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2595 * config/arc/arc.md (add_shift): New pattern.
2596 (add_shift2): Likewise.
2597 (sub_shift): Likewise.
2598 (sub_shift_cmp0_noout): Likewise.
2599 (compare_si_ashiftsi): Likewise.
2600 (xbfu_cmp0_noout): New combine pattern.
2601 (xbfu_cmp0"): Likewise.
2602 (movsi_set_cc_insn): Place the predicable variant first.
2603 (commutative_binary_cmp0_noout): Remove clobber.
2604 (commutative_binary_cmp0): New pattern.
2605 (noncommutative_binary_cmp0): Likewise.
2606 (noncommutative_binary_cmp0_noout): Likewise.
2607 (noncommutative_binary_comparison_result_used): Removed.
2608 (rsub_cmp0): New pattern.
2609 (rsub_cmp0_noout): Likewise.
2610 (extzvsi): Changed, keep only meaningful variants.
2611 (SQH, SEZ): New iterators.
2612 (SQH_postfix): New mode attribute.
2613 (SEZ_prefix): New code attribute.
2614 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2615 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2616 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2617 of numerical value.
2618 (noncommutative_operator): Check the availability of barrel
2619 shifter option.
2621 2018-04-09 Richard Biener <rguenther@suse.de>
2623 PR tree-optimization/85284
2624 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2625 Only use the niter constraining form of simple_iv when the exit
2626 is always executed.
2628 2018-04-09 Tom de Vries <tom@codesourcery.com>
2630 PR target/84041
2631 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2632 (define_expand "*memory_barrier"): New define_expand.
2633 (define_insn "memory_barrier"): New insn.
2635 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2637 PR rtl-optimization/80463
2638 PR rtl-optimization/83972
2639 PR rtl-optimization/83480
2641 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2642 correct producer for the insn.
2643 (tidy_control_flow): Fixup seqnos in case of debug insns.
2645 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2647 PR rtl-optimization/83913
2649 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2650 different sched-times when merging exprs.
2652 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2654 PR rtl-optimization/83962
2656 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2657 tidy_fallthru_edge and tidy_control_flow.
2659 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2661 PR rtl-optimization/83530
2663 * sel-sched.c (force_next_insn): New global variable.
2664 (remove_insn_for_debug): When force_next_insn is true, also leave only
2665 next insn in the ready list.
2666 (sel_sched_region): When the region wasn't scheduled, make another pass
2667 over it with force_next_insn set to 1.
2669 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2671 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2672 into tm_file.
2673 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2674 for interrupt control.
2675 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2676 functions for interrupt control.
2677 * config/nds32/nds32-intrinsic.md: Likewise.
2678 * config/nds32/nds32_intrinsic.h: Likewise.
2679 * config/nds32/nds32.h (nds32_builtins): Likewise.
2681 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2683 * config/nds32/nds32.c (nds32_init_machine_status,
2684 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2685 strict_aligned_p field.
2686 (nds32_expand_to_rtl_hook): New function.
2687 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2688 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2690 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2691 Chung-Ju Wu <jasonwucj@gmail.com>
2693 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2694 * config/nds32/nds32-n7.md: New file.
2695 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2696 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2697 pipeline.
2698 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2699 * config/nds32/nds32.md (pipeline_model): Add n7.
2700 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2701 * config/nds32/pipelines.md: Include n7 settings.
2703 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2704 Chung-Ju Wu <jasonwucj@gmail.com>
2706 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2707 * config/nds32/nds32-e8.md: New file.
2708 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2709 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2710 pipeline.
2711 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2712 * config/nds32/nds32.md (pipeline_model): Add e8.
2713 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2714 * config/nds32/pipelines.md: Include e8 settings.
2716 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2717 Chung-Ju Wu <jasonwucj@gmail.com>
2719 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2720 * config/nds32/nds32-n8.md: New file.
2721 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2722 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2723 pipeline.
2724 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2725 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2726 * config/nds32/nds32.md (pipeline_model): Add n8.
2727 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2728 * config/nds32/pipelines.md: Include n8 settings.
2730 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2731 Chung-Ju Wu <jasonwucj@gmail.com>
2733 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2734 * config/nds32/nds32-n9-2r1w.md: New file.
2735 * config/nds32/nds32-n9-3r2w.md: New file.
2736 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2737 nds32_register_ports): New or modify for cpu n9.
2738 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2739 pipeline.
2740 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2741 * config/nds32/nds32-utils.c: New file.
2742 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2743 TARGET_MUL_SLOW): Define.
2744 * config/nds32/nds32.md (pipeline_model): New attribute.
2745 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2746 New options that support cpu n9.
2747 * config/nds32/pipelines.md: Include n9 settings.
2748 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2750 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2752 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2753 information if necessary.
2754 (output_cond_branch_compare_zero): Likewise.
2755 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2756 (nds32_target_alignment): Refine for alignment.
2757 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2758 (FUNCTION_BOUNDARY): Modify.
2759 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2760 align case.
2761 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2763 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2765 * config/nds32/constants.md (unspec_volatile_element): Add values for
2766 TLB operation and data prefetch.
2767 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2768 functions for TLB operation and data prefetch.
2769 * config/nds32/nds32-intrinsic.md: Likewise.
2770 * config/nds32/nds32_intrinsic.h: Likewise.
2771 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2772 (nds32_print_operand): Likewise.
2773 * config/nds32/nds32.h (nds32_builtins): Likewise.
2775 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2776 Andrew Pinski <pinsika@gcc.gnu.org>
2778 PR middle-end/82976
2779 * match.pd: Use constant_boolean_node of correct type instead of
2780 boolean_true_node or boolean_false_node for simplifying
2781 pointer comparisons to zero.
2783 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2785 PR tree-optimization/80021
2786 * tree.c (verify_type_variant): Make error call in verify_variant_match
2787 translatable and remove final full stop.
2789 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2791 * config/nds32/constants.md (unspec_volatile_element): Add
2792 UNSPEC_VOLATILE_EH_RETURN.
2793 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2794 nds32_output_stack_pop): Support dwarf exception handling process.
2795 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2796 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2797 exception handling process.
2798 (nds32_compute_stack_frame): Likewise.
2799 (nds32_return_addr_rtx): Likewise.
2800 (nds32_initial_elimination_offset): Likewise.
2801 (nds32_expand_prologue): Likewise.
2802 (nds32_expand_epilogue): Likewise.
2803 (nds32_dynamic_chain_address): New function.
2804 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2805 exception handling.
2806 (DYNAMIC_CHAIN_ADDRESS): Define.
2807 (EH_RETURN_DATA_REGNO): Define.
2808 (EH_RETURN_STACKADJ_RTX): Define.
2809 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2810 patterns for dwarf exception handling.
2812 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2814 * config/nds32/nds32.h: Clean up obsolete macros.
2816 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2818 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2819 Add enum values for particular instructions.
2820 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2821 particular intrinsic functions.
2822 * config/nds32/nds32-intrinsic.md: Likewise.
2823 * config/nds32/nds32_intrinsic.h: Likewise.
2824 * config/nds32/nds32.h (nds32_builtins): Likewise.
2825 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2826 (btst, ave): New patterns for particular instructions.
2828 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2830 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2831 Add enum values for atomic load/store and memory sync.
2832 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2833 and memory sync.
2834 * config/nds32/nds32-intrinsic.md: Likewise.
2835 * config/nds32/nds32_intrinsic.h: Likewise.
2836 * config/nds32/nds32.h (nds32_builtins): Likewise.
2838 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2840 PR tree-optimization/85257
2841 * fold-const.c (native_encode_vector): If not all elts could fit
2842 and off is -1, return 0 rather than offset.
2843 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2844 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2845 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2846 adjust buffer in native_interpret_expr call.
2848 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2850 * config/nds32/constants.md (unspec_volatile_element): Add cache
2851 control enum values.
2852 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2853 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2854 * config/nds32/nds32.c (nds32_cctl_names): New.
2855 (nds32_print_operand): Handle cache control register names.
2856 * config/nds32/nds32.h (nds32_builtins): New enum values.
2857 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2858 macros.
2859 * config/nds32/nds32.md (type): Add mmu.
2860 * config/nds32/pipelines.md (simple_insn): Add mmu.
2862 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2864 * config/nds32/nds32.md (type): Remove call.
2865 * config/nds32/pipelines.md (simple_insn): Likewise.
2867 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2869 * config/nds32/constants.md (unspec_volatile_element): Add
2870 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2871 UNSPEC_VOLATILE_FMFCFG.
2872 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2873 description for fmfcfg and fmfcsr.
2874 (bdesc_1arg): Add fmtcsr.
2875 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2876 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2877 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2878 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2879 unspec_fmfcfg): New patterns.
2880 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2881 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2882 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2883 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2884 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2885 __nds32__fmfcfg): Define.
2887 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2889 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2890 intrinsic register names.
2891 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2892 intrinsic register enum values and macros.
2894 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2896 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2897 for load/store addressing form.
2898 (nds32_print_operand_address): Likewise.
2900 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2902 PR target/85196
2903 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2904 based on LABEL_REF. Remove useless assertion.
2905 (pic_address_needs_scratch): Fix formatting.
2906 (sparc_legitimize_pic_address): Minor tweaks.
2907 (sparc_delegitimize_address): Adjust assertion accordingly.
2908 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2909 into symbolic_operand.
2910 (movsi_high_pic_label_ref): Likewise.
2911 (movsi_lo_sum_pic_label_ref): Likewise.
2912 (movdi_pic_label_ref): Likewise.
2913 (movdi_high_pic_label_ref): Likewise.
2914 (movdi_lo_sum_pic_label_ref): Likewise.
2916 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2918 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2919 custom LIB_SPEC setup.
2921 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2922 Kito Cheng <kito.cheng@gmail.com>
2924 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2925 * config/riscv/freebsd.h: New.
2927 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2929 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2930 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2931 file.
2933 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2934 Kito Cheng <kito.cheng@gmail.com>
2936 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2937 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2938 * config/nds32/nds32-protos.h (nds32_output_call,
2939 nds32_output_return): Declare.
2940 * config/nds32/nds32.md: Refine all the call and return patterns.
2942 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2944 PR debug/85252
2945 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2946 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2948 PR rtl-optimization/84872
2949 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2950 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2951 EDGE_CROSSING edge.
2953 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2955 * expr.c (copy_blkmode_to_reg): Revert 254862.
2956 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2958 2018-04-06 Richard Biener <rguenther@suse.de>
2960 PR middle-end/85244
2961 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2962 after seeing a component reference with an adjacent field. Treat
2963 refs to arrays at struct end of external decls similar to
2964 refs to unconstrained commons.
2966 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2968 PR sanitizer/85213
2969 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2970 look through SAVE_EXPRs with non-side-effects argument. Adjust
2971 recursive calls.
2972 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2973 save_p here.
2975 2018-04-06 Richard Biener <rguenther@suse.de>
2977 PR middle-end/85180
2978 * alias.c (find_base_term): New wrapper around find_base_term
2979 unwinding CSELIB_VAL_PTR changes.
2980 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2981 recursion.
2983 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2985 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2986 instructions.
2987 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2988 constant definitions.
2989 ("nop"): lr 0,0 -> nopr r0
2990 ("nop_lr0", "nop_lr1"): New insn definitions.
2992 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2994 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2995 NDS32_V3PUSH_AVAILABLE_P macro.
2997 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2998 Chung-Ju Wu <jasonwucj@gmail.com>
3000 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3001 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3002 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3003 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3004 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3005 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3006 * config/nds32/constraints.md: New constraints and checking for hard
3007 float configuration.
3008 * config/nds32/iterators.md: New mode iterator and attribute for hard
3009 float configuration.
3010 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3011 patterns.
3012 * config/nds32/nds32-fpu.md: New file.
3013 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3014 deal with hard float code generation.
3015 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3016 ARCH_V3S.
3017 (abi_type, float_reg_number): New enum type.
3018 * config/nds32/nds32-predicates.c: New predicates for hard float.
3019 * config/nds32/nds32-protos.h: Declare functions for hard float.
3020 * config/nds32/nds32.c: Implementation for hard float configuration.
3021 * config/nds32/nds32.h: Definitions for hard float configuration.
3022 * config/nds32/nds32.md: Include hard float machine description and
3023 modify patterns for hard float configuration.
3024 * config/nds32/nds32.opt: New options for hard float configuration.
3025 * config/nds32/predicates.md: New predicates for hard float
3026 configuration.
3028 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3030 * common/config/nds32/nds32-common.c
3031 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3033 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3035 PR middle-end/85195
3036 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3037 CONSTRUCTOR_ELT (ctor, ...)->value.
3039 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3041 PR target/85193
3042 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3044 2018-04-05 Tom de Vries <tom@codesourcery.com>
3046 PR target/85204
3047 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3048 cond jump.
3050 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3051 Kito Cheng <kito.cheng@gmail.com>
3053 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3054 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3055 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3056 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3058 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3059 Kito Cheng <kito.cheng@gmail.com>
3061 * config/nds32/constraints.md (Ufe): New memory constraint.
3062 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3063 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3064 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3065 operands.
3066 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3067 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3069 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3071 * config/nds32/nds32.md: Use optimize_size in the condition for
3072 alu-shift instructions.
3074 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3076 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3078 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3080 * config/nds32/nds32.md (negsi2): Refine pattern.
3082 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3083 Chung-Ju Wu <jasonwucj@gmail.com>
3085 * config/nds32/iterators.md (shift_rotate): New code iterator.
3086 (shift): New code attribute.
3087 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3088 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3089 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3090 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3091 bit-wise operations.
3092 (andsi3, *andsi3): Ditto.
3093 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3094 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3095 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3096 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3097 nds32_ior_operand, nds32_xor_operand): New predicates.
3099 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3101 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3102 (addsi3, subsi3): ... this.
3104 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3106 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3108 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3110 * config/nds32/nds32.md: Adjust indention.
3112 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3114 * config/nds32/nds32.md (feature): New attribute.
3116 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3118 * config/nds32/nds32.md (subtype): New attribute.
3120 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3122 PR target/85203
3123 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3124 expansion to perform a bitwise AND of the argument followed by a
3125 boolean negation of the result.
3127 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3129 PR rtl-optimization/84878
3130 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3131 the basic block. Assert the use reference is not artificial and that
3132 it has an associated insn.
3134 2018-04-04 Michael Matz <matz@suse.de>
3136 * builtins.c (compute_objsize): Pass correct operand
3137 to array_at_struct_end_p.
3139 2018-04-04 Richard Biener <rguenther@suse.de>
3141 PR lto/85176
3142 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3143 from contexts for DINFO_LEVEL_TERSE and below.
3145 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3147 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3148 resiter_operand condition.
3149 * config/nds32/nds32.md (*move<mode>): Ditto.
3151 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3152 Monk Chiang <sh.chiang04@gmail.com>
3154 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3156 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3158 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3160 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3161 Kito Cheng <kito.cheng@gmail.com>
3163 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3164 nds32_cond_code_str, output_cond_branch,
3165 output_cond_branch_compare_zero, nds32_expand_cbranch,
3166 nds32_expand_cstore, nds32_expand_movcc,
3167 nds32_output_cbranchsi4_equality_zero,
3168 nds32_output_cbranchsi4_equality_reg,
3169 nds32_output_cbranchsi4_equality_reg_or_const_int,
3170 nds32_output_cbranchsi4_greater_less_zero: New functions.
3171 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3172 nds32_expand_cstore, nds32_expand_movcc,
3173 nds32_output_cbranchsi4_equality_zero,
3174 nds32_output_cbranchsi4_equality_reg,
3175 nds32_output_cbranchsi4_equality_reg_or_const_int,
3176 nds32_output_cbranchsi4_greater_less_zero): Declare.
3177 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3178 nds32_rimm11s_operand): New predicates.
3179 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3180 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3181 patterns.
3183 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3185 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3186 * config/nds32/nds32.md: Ditto.
3187 * config/nds32/pipelines.md: Ditto.
3189 2018-04-04 Richard Biener <rguenther@suse.de>
3191 PR tree-optimization/85168
3192 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3193 propagating abnormals.
3195 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3197 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3199 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3200 Kito Cheng <kito.cheng@gmail.com>
3202 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3203 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3204 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3205 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3206 * config/nds32/nds32.md (sibcall_internal): New.
3207 (sibcall_register): Remove.
3208 (sibcall_immediate): Remove.
3209 (sibcall_value_internal): New.
3210 (sibcall_value_register): Remove.
3211 (sibcall_value_immediate): Remove.
3212 * config/nds32/predicates.md (nds32_general_register_operand): New.
3213 (nds32_call_address_operand): New.
3215 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3217 PR rtl-optimization/85167
3218 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3219 bb_defs if *split_p, instead preinitialize it to NULL.
3221 PR tree-optimization/85156
3222 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3223 evaluating the argument multiple times.
3225 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3227 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3228 than vector.
3229 (_mm_cvtpd_ps): Likewise.
3230 (_mm_cvttpd_epi32): Likewise.
3231 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3232 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3233 vector, pixel, and bool following altivec.h include.
3235 2018-04-03 Martin Sebor <msebor@redhat.com>
3237 * doc/extend.texi (Common Function Attributes): Clarify.
3238 (const attribute): Likewise.
3239 (pure attribute): Likewise.
3241 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3243 PR target/85169
3244 * config/i386/i386.c (ix86_expand_vector_set): Use
3245 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3247 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3249 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3250 instructions when changing rounding bits to preserve precision bits
3251 in the x87 control word.
3253 2018-04-03 Martin Liska <mliska@suse.cz>
3255 PR tree-optimization/82491
3256 * rtl.h (strip_offset_and_add): Replace += suboffset with
3257 poly_uint64 () + suboffset.
3259 2018-03-29 Martin Liska <mliska@suse.cz>
3260 Martin Jambor <mjambor@suse.cz>
3262 PR ipa/84947
3263 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3264 param_type is not an integral or pointer type.
3266 2018-04-03 Richard Biener <rguenther@suse.de>
3268 * sese.h (recompute_all_dominators): Remove.
3270 2018-04-02 Martin Sebor <msebor@redhat.com>
3272 * doc/invoke.texi (-Wrestrict): Fix typos.
3274 2018-04-02 Jim Wilson <jimw@sifive.com>
3276 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3277 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3278 (<optab>di3, <optab>si3_extend): Likewise.
3279 (<optab>si3_mask, <optab>si3_mask_1): New.
3280 (<optab>di3_mask, <optab>di3_mask_1): New.
3281 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3282 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3283 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3285 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3287 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3288 example.
3290 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3292 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3293 (nds32_canonicalize_comparison): New function.
3295 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3296 Kito Cheng <kito.cheng@gmail.com>
3297 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3299 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3300 * config/nds32/constants.md (unspec_volatile_element): Add
3301 UNSPEC_VOLATILE_RELAX_GROUP.
3302 * config/nds32/nds32-relax-opt.c: New file.
3303 * config/nds32/nds32-predicates.c
3304 (nds32_symbol_load_store_p): New function.
3305 * config/nds32/nds32-protos.h
3306 (nds32_symbol_load_store_p): Declare function.
3307 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3308 * config/nds32/nds32.c
3309 (nds32_register_pass): New function to register pass.
3310 (nds32_register_passes): New function to register passes.
3311 * config/nds32/nds32.md (relax_group): New pattern.
3312 * config/nds32/nds32.opt (mrelax-hint): New option.
3313 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3315 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3317 * config/nds32/t-nds32: Modify files dependency.
3319 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3321 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3322 (PROFILE_HOOK): Define its implementation.
3324 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3326 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3327 type and 32-bit size.
3329 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3331 PR middle-end/85090
3332 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3333 (V_128_256): New mode iterator.
3334 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3335 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3336 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3337 of V.
3338 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3339 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3341 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3343 PR target/83315
3344 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3345 NaN inputs correctly.
3347 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3349 PR target/80546
3350 * config/rs6000/vsx.md (??r): New mode attribute.
3351 (*vsx_mov<mode>_64bit): Use it.
3352 (*vsx_mov<mode>_32bit): Likewise.
3354 2018-03-30 Martin Sebor <msebor@redhat.com>
3356 PR tree-optimization/84818
3357 * builtins.c (check_access): Use warning_n.
3359 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3361 PR target/83822
3362 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3363 condition.
3364 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3365 condition.
3367 2018-03-30 Julia Koval <julia.koval@intel.com>
3369 PR target/84413
3370 * x86-tune.def (movx, partial_reg_dependency): Enable for
3371 m_SKYLAKE_AVX512.
3373 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3375 PR inline-asm/84985
3376 * lra-constraints.c (process_alt_operands): Move setting
3377 this_alternative_matches below.
3379 2018-03-29 Martin Liska <mliska@suse.cz>
3381 PR lto/84995.
3382 * doc/invoke.texi: Document how LTO works with debug info.
3383 Describe auto-load support of binutils. Mention 'x86-64'
3384 as valid option value of -march option.
3386 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3388 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3390 PR c/85094
3391 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3392 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3393 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3394 checking.
3396 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3398 PR target/84912
3399 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3400 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3401 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3402 for RS6000_BTM_POWERPC64.
3403 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3404 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3405 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3406 definition.
3407 (DIVDE): Use it.
3408 (DIVDEU): Likewise.
3410 2018-03-28 Carl Love <cel@us.ibm.com>
3412 Revert
3413 2017-09-27 Carl Love <cel@us.ibm.com>
3415 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3416 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3417 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3418 fctiw instruction.
3420 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3422 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3423 instead of __vector bool.
3424 (_mm_max_pu8): Likewise.
3425 (_mm_min_pi16): Likewise.
3427 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3429 PR target/84912
3430 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3431 (DIVWEUO): Likewise.
3432 (DIVDEO): Likewise.
3433 (DIVDEUO): Likewise.
3434 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3435 DIVWEUO and DIVDEUO.
3436 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3437 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3438 (div_extend): Likewise.
3439 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3440 builtin function.
3441 (__builtin_divweuo): Likewise.
3442 (__builtin_divdeo): Likewise.
3443 (__builtin_divdeuo): Likewise.
3445 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3447 PR target/85095
3448 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3449 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3451 PR tree-optimization/82004
3452 * gimple-match-head.c (optimize_pow_to_exp): New function.
3453 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3454 Don't fold to exp if optimize_pow_to_exp is false.
3456 2018-03-28 Martin Liska <mliska@suse.cz>
3458 PR other/84819
3459 * calls.c (initialize_argument_information): Fix trailing space.
3460 * common.opt: Fix typo and provide better explanation for
3461 -fsanitize-coverage option.
3462 * config/i386/i386.opt: Fix typo.
3464 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3465 Martin Liska <mliska@suse.cz>
3467 PR sanitizer/85081
3468 * gimplify.c (asan_poison_variable): Don't do the check for
3469 gimplify_omp_ctxp here.
3470 (gimplify_decl_expr): Do it here.
3471 (gimplify_target_expr): Likewise.
3473 2018-03-28 Martin Liska <mliska@suse.cz>
3475 PR target/84988
3476 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3477 chkp_type_bounds_count if MPX is not enabled.
3479 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3481 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3483 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3485 PR target/84914
3486 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3487 function to create the function decl for complex long double
3488 multiply and divide for -mabi=ieeelongdouble.
3489 (init_float128_ieee): Call it.
3491 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3493 PR target/85044
3494 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3495 -fcf-protection=branch -mibt.
3496 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3498 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3500 PR target/81863
3501 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3503 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3505 PR target/85056
3506 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3507 extern array declarations.
3509 2018-03-27 Richard Biener <rguenther@suse.de>
3511 PR middle-end/84067
3512 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3513 explicit single_use checks.
3515 2018-03-27 Richard Biener <rguenther@suse.de>
3517 PR tree-optimization/85082
3518 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3519 Valueize the VUSE.
3521 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3523 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3524 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3525 Turn on fasynchronous-unwind-tables and funwind-tables.
3527 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3529 PR target/85073
3530 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3531 (*bmi_blsr_<mode>_ccz): Ditto.
3533 2018-03-26 Tom de Vries <tom@codesourcery.com>
3535 PR tree-optimization/85063
3536 * omp-general.c (offloading_function_p): New function. Factor out
3537 of ...
3538 * omp-offload.c (pass_omp_target_link::gate): ... here.
3539 * omp-general.h (offloading_function_p): Declare.
3540 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3541 with attribute omp declare target for offloading functions.
3543 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3545 PR tree-optimization/84005
3546 * tree-data-ref.h (get_base_for_alignment): Declare.
3547 * tree-data-ref.c (get_base_for_alignment_1): New function.
3548 (get_base_for_alignment): Likewise.
3549 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3550 get_base_for_alignment to find a suitable base object, instead
3551 of always using drb->base_address.
3553 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3555 PR inline-asm/85022
3556 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3557 known size by default.
3559 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3561 PR inline-asm/85030
3562 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3563 and non BLKmode operands.
3565 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3567 PR target/85026
3568 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3569 Clean up attributes.
3571 2018-03-23 Richard Biener <rguenther@suse.de>
3573 PR debug/85020
3574 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3575 we are going to emit early debug for LTO.
3577 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3579 PR inline-asm/85034
3580 * function.c (match_asm_constraints_1): Don't optimize if input
3581 doesn't satisfy general_operand predicate for output's mode.
3583 PR inline-asm/85022
3584 * alias.c (write_dependence_p): Don't require for x_canonicalized
3585 non-VOIDmode if x has VOIDmode.
3587 PR sanitizer/85029
3588 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3589 just don't try to optimize it rather than assert it never happens.
3591 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3593 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3594 macro expansions for definition of ST_INTERNAL_<mode> and
3595 LD_INTERNAL_<mode> builtins.
3596 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3597 Remove prototype.
3598 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3599 function.
3600 (altivec_expand_st_builtin): Likewise.
3601 (altivec_expand_builtin): Remove calls to deleted functions.
3602 (rs6000_address_for_altivec): Delete this function.
3603 * config/rs6000/vector.md: Remove expands for
3604 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3606 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3608 PR target/84826
3609 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3610 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3611 re-computing once computed.
3612 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3613 (arm_init_machine_status): Initialize
3614 machine->static_chain_stack_bytes.
3616 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3618 PR target/84760
3619 * doc/extend.texi: Add four new prototypes for vec_ld.
3620 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3621 definitions for more logical presentation.
3622 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3623 entries for V1TI variants of __builtin_altivec_ld builtin.
3624 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3625 handling of V1TI variant of LVX icode pattern.
3626 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3627 (rs6000_gimple_fold_builtin): Likewise.
3628 (altivec_init_builtins): Add code to define
3629 __builtin_altivec_lvx_v1ti function.
3631 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3633 PR inline-asm/84941
3634 * function.c (match_asm_constraints_1): Don't do the optimization
3635 if input isn't a REG, SUBREG, MEM or constant.
3637 2018-03-22 Tom de Vries <tom@codesourcery.com>
3639 PR tree-optimization/84956
3640 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3641 bb_has_abnormal_pred.
3643 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3645 PR sanitizer/85018
3646 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3647 DECL_INITIAL (decl) to decl at the end.
3648 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3649 adjust the comment.
3651 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3653 * doc/extend.texi (__builtin_tgmath): Document when complex
3654 integer types are treated as _Complex _Float64.
3656 2018-03-21 Tom de Vries <tom@codesourcery.com>
3658 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3660 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3662 PR tree-optimization/84960
3663 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3664 if it is ENTRY block, move them into single succ of ENTRY in that case.
3666 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3668 PR tree-optimization/84811
3669 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3670 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3671 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3672 (known_subrange_p): Update accordingly. Cast each value involved
3673 in the size comparison, rather than casting the result of the
3674 subtraction.
3676 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3678 PR tree-optimization/84982
3679 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3680 by flipping the least significant bit rather than all bits from
3681 bitpos to bitpos + bitsize - 1.
3683 2018-03-21 Nathan Sidwell <nathan@acm.org>
3685 * doc/extend.texi (Deprecated Features): Remove mention of
3686 long-deleted deprecations.
3688 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3690 PR jit/84288
3691 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3692 * configure: Regenerate.
3694 2018-03-21 Tom de Vries <tom@codesourcery.com>
3696 PR tree-optimization/83126
3697 * tree-parloops.c (num_phis): New function.
3698 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3700 2018-03-21 Nathan Sidwell <nathan@acm.org>
3702 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3703 mention anon-struct/union members and trailing attributes.
3705 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3707 PR tree-optimization/84969
3708 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3709 builtin memset partitions if they set different rhs values.
3711 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3713 PR rtl-optimization/84989
3714 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3715 VEC_DUPLICATE with scalar result mode.
3717 2018-03-21 Martin Liska <mliska@suse.cz>
3719 PR ipa/84963
3720 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3721 not intended return statement.
3723 2018-03-21 Martin Liska <mliska@suse.cz>
3725 PR target/84988
3726 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3727 (chkp_find_bound_slots_1): Limit number of iterations.
3729 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3731 PR target/84838
3732 * Minor grammar fixes for x86 options.
3734 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3736 PR debug/84875
3737 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3738 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3740 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3742 PR target/83789
3743 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3744 (altivec_lvx_<mode>_1op): Likewise.
3745 (altivec_stvx_<mode>_2op): Likewise.
3746 (altivec_stvx_<mode>_1op): Likewise.
3747 (altivec_lvx_<VM2:mode>): New define_expand.
3748 (altivec_stvx_<VM2:mode>): Likewise.
3749 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3750 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3751 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3752 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3753 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3754 (rs6000_gen_lvx): Likewise.
3755 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3756 (altivec_expand_stv_builtin): Likewise.
3757 (altivec_expand_builtin): Likewise.
3758 * config/rs6000/vector.md: Likewise.
3760 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3762 PR target/82518
3763 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3764 BYTES_BIG_ENDIAN.
3766 2018-03-20 Richard Biener <rguenther@suse.de>
3768 PR target/84986
3769 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3770 sign-conversions as zero, fall back to standard scalar_stmt
3771 cost for the rest.
3773 2018-03-20 Martin Liska <mliska@suse.cz>
3775 PR ipa/84825
3776 * predict.c (rebuild_frequencies): Handle case when we have
3777 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3779 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3781 PR target/84990
3782 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3783 flag_section_anchors.
3784 * varasm.c (use_blocks_for_decl_p): Remove hack for
3785 dw2_force_const_mem.
3787 PR target/84845
3788 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3789 to ...
3790 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3791 be created, use lowpart_subreg of operands[0] rather than operands[0]
3792 itself.
3793 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3794 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3795 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3796 and n constraint instead of aarch64_shift_imm_di and Usd.
3797 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3798 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3800 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3802 PR target/82989
3803 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3804 to favor GPR over NEON registers.
3805 (<shift>di3_neon): Likewise.
3807 2018-03-20 Tom de Vries <tom@codesourcery.com>
3809 PR target/84952
3810 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3811 (nvptx_process_pars): Emit bar.sync asap and alap.
3813 2018-03-20 Tom de Vries <tom@codesourcery.com>
3815 PR target/84954
3816 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3817 seen_label if seen_label is already set.
3819 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3821 PR target/84945
3822 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3823 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3824 Use 1U instead of 1. Formatting fixes.
3826 PR c/84953
3827 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3828 instead of TREE_TYPE (s1) for the return value.
3830 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3832 PR tree-optimization/84946
3833 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3834 bitsize + bitsize in poly_uint64 rather than poly_int64.
3836 PR sanitizer/78651
3837 * dwarf2asm.c: Include fold-const.c.
3838 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3839 of decl rather than decl itself.
3841 PR rtl-optimization/84643
3842 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3844 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3846 PR sanitizer/78651
3847 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3848 calling assemble_variable.
3850 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3852 PR target/81647
3853 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3854 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3856 2018-03-19 Jim Wilson <jimw@sifive.com>
3858 PR bootstrap/84856
3859 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3860 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3861 (riscv_first_stack_step): Likewise.
3862 (riscv_option_override): Use STACK_BOUNDARY instead of
3863 MIN_STACK_BOUNDARY.
3864 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3865 MIN_STACK_BOUNDARY.
3866 (BIGGEST_ALIGNMENT): Set to 128.
3867 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3868 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3869 STACK_BOUNDARY.
3871 2018-03-19 Richard Biener <rguenther@suse.de>
3873 PR tree-optimization/84933
3874 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3875 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3877 2018-03-19 Richard Biener <rguenther@suse.de>
3879 PR tree-optimization/84859
3880 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3881 (cond_if_else_store_replacement): Perform sinking operation on
3882 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3883 Generalize what a BB with a single eligible store is.
3885 2018-03-19 Richard Biener <rguenther@suse.de>
3887 PR tree-optimization/84929
3888 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3889 chrec_is_positive against non-chrec arg.
3891 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3893 PR target/84711
3894 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3896 2018-03-18 Martin Liska <mliska@suse.cz>
3898 PR rtl-optimization/84635
3899 * regrename.c (build_def_use): Use matches_mode only when
3900 matches >= 0.
3902 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3904 PR tree-optimization/84913
3905 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3906 vectorize chains of COND_EXPRs.
3908 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3912 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3914 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3916 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3918 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3920 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3921 Kito Cheng <kito.cheng@gmail.com>
3923 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3924 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3925 (nds32_adjust_reg_alloc_order): New function.
3926 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3928 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3930 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3931 nds32_print_operand, nds32_print_operand_address): Use
3932 HOST_WIDE_INT_PRINT_DEC instead.
3934 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3936 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3938 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3940 PR target/84902
3941 * config/i386/i386.c (initial_ix86_tune_features,
3942 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3943 unsigned long long.
3944 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3945 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3946 rather than 1u << ix86_tune. Formatting fix.
3947 (ix86_option_override_internal): Change ix86_arch_mask from
3948 unsigned int to unsigned HOST_WIDE_INT, initialize to
3949 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3950 (ix86_function_specific_restore): Likewise.
3952 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3954 PR target/84899
3955 * postreload.c (reload_combine_recognize_pattern): Perform
3956 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3957 truncate_int_for_mode the result for the destination's mode.
3959 PR c/84909
3960 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3961 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3962 Likewise.
3963 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3964 Likewise.
3966 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3968 PR target/84876
3969 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3970 regno_allocno_class_array and sorted_pseudos.
3971 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3972 insns where regno is used.
3974 2018-03-16 Martin Liska <mliska@suse.cz>
3976 PR ipa/84833
3977 * multiple_target.c (create_dispatcher_calls): Redirect
3978 reference in the symbol table.
3980 2018-03-16 Martin Liska <mliska@suse.cz>
3982 PR ipa/84722
3983 * multiple_target.c (create_dispatcher_calls): Redirect also
3984 an alias.
3986 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3988 PR c++/79937
3989 PR c++/82410
3990 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3991 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3992 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3994 2018-03-16 Julia Koval <julia.koval@intel.com>
3996 * doc/invoke.texi (Skylake Server): Add CLWB.
3997 Cannonlake): Remove CLWB.
3999 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4001 PR tree-optimization/84841
4002 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4003 1 << 3.
4004 (FLOAT_ONE_CONST_TYPE): Define.
4005 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4006 (sort_by_operand_rank): Put entries with higher constant_type last
4007 rather than first to match comments.
4009 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4011 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4012 split predicate.
4014 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4016 PR c++/79085
4017 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4018 check and use address of target always.
4020 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4022 PR target/84574
4023 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4024 (indirect_thunk_bnd_needed): Likewise.
4025 (indirect_thunks_used): Likewise.
4026 (indirect_thunks_bnd_used): Likewise.
4027 (indirect_return_needed): New.
4028 (indirect_return_bnd_needed): Likewise.
4029 (output_indirect_thunk_function): Add a bool argument for
4030 function return.
4031 (output_indirect_thunk_function): Don't generate alias for
4032 function return thunk.
4033 (ix86_code_end): Call output_indirect_thunk_function to generate
4034 function return thunks.
4035 (ix86_output_function_return): Set indirect_return_bnd_needed
4036 and indirect_return_needed instead of indirect_thunk_bnd_needed
4037 and indirect_thunk_needed.
4039 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4041 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4042 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4043 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4045 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4046 Paul Hua <paul.hua.gm@gmail.com>
4048 PR c/84852
4049 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4051 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4053 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4054 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4055 resp. SFmode cases.
4057 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4059 PR target/84711
4060 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4061 instead of GET_MODE_SIZE when comparing Units.
4063 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4065 PR target/68256
4066 * varasm.c (hash_section): Return an unchangeble hash value
4067 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4068 Return !aarch64_can_use_per_function_literal_pools_p ().
4070 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4072 PR target/84860
4073 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4074 rather than address of cmode as last argument to prepare_cmp_insn.
4076 2018-03-15 Julia Koval <julia.koval@intel.com>
4078 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4079 F_AVX512VNNI, F_AVX512BITALG): New.
4081 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4083 PR target/83451
4084 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4085 insn for floating-point loads and stores.
4087 2018-03-14 Carl Love <cel@us.ibm.com>
4089 * config/rs6000/rs6000-c.c: Add macro definitions for
4090 ALTIVEC_BUILTIN_VEC_PERMXOR.
4091 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4092 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4093 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4094 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4095 UNSPEC_VPERMXOR.
4096 * config/doc/extend.texi: Add prototypes for vec_permxor.
4098 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4100 PR c/84852
4101 * diagnostic-show-locus.c (class layout_point): Convert m_line
4102 from int to linenum_type.
4103 (line_span::comparator): Use linenum "compare" function when
4104 comparing line numbers.
4105 (test_line_span): New function.
4106 (layout_range::contains_point): Convert param "row" from int to
4107 linenum_type.
4108 (layout_range::intersects_line_p): Likewise.
4109 (layout::will_show_line_p): Likewise.
4110 (layout::print_source_line): Likewise.
4111 (layout::should_print_annotation_line_p): Likewise.
4112 (layout::print_annotation_line): Likewise.
4113 (layout::print_leading_fixits): Likewise.
4114 (layout::annotation_line_showed_range_p): Likewise.
4115 (struct line_corrections): Likewise for field m_row.
4116 (line_corrections::line_corrections): Likewise for param "row".
4117 (layout::print_trailing_fixits): Likewise.
4118 (layout::get_state_at_point): Likewise.
4119 (layout::get_x_bound_for_row): Likewise.
4120 (layout::print_line): Likewise.
4121 (diagnostic_show_locus): Likewise for locals "last_line" and
4122 "row".
4123 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4124 * input.c (selftest::test_linenum_comparisons): New function.
4125 (selftest::input_c_tests): Call it.
4126 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4127 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4128 * selftest.h (ASSERT_GT): New macro.
4129 (ASSERT_GT_AT): New macro.
4130 (ASSERT_LT): New macro.
4131 (ASSERT_LT_AT): New macro.
4133 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4135 PR rtl-optimization/84780
4136 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4138 2018-03-14 Martin Liska <mliska@suse.cz>
4140 * tree.c (record_node_allocation_statistics): Use
4141 get_stats_node_kind.
4142 (get_stats_node_kind): New function extracted from
4143 record_node_allocation_statistics.
4144 (free_node): Use get_stats_node_kind.
4146 2018-03-14 Richard Biener <rguenther@suse.de>
4148 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4149 that the value-set of ANTIC_IN doesn't grow.
4151 Revert
4152 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4153 member.
4154 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4155 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4157 2018-03-14 Julia Koval <julia.koval@intel.com>
4159 * config.gcc (icelake-client, icelake-server): New.
4160 (icelake): Remove.
4161 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4162 (initial_ix86_arch_features): Ditto.
4163 (PTA_SKYLAKE): Add SGX.
4164 (PTA_ICELAKE): Remove.
4165 (PTA_ICELAKE_CLIENT): New.
4166 (PTA_ICELAKE_SERVER): New.
4167 (ix86_option_override_internal): Split up icelake on icelake client and
4168 icelake server.
4169 (get_builtin_code_for_version): Ditto.
4170 (fold_builtin_cpu): Ditto.
4171 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4172 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4173 * config/i386/i386.h (processor_type): Ditto.
4174 * doc/invoke.texi: Ditto.
4176 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4178 PR sanitizer/83392
4179 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4180 INTEGER_CST offset, add it together with bitpos / 8 and
4181 sign extend based on POINTER_SIZE.
4183 PR target/84844
4184 Revert
4185 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4187 PR target/78090
4188 * config/i386/constraints.md (Yc): New register constraint.
4189 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4190 Use Yc constraint for alternative 2 of operand 0. Remove
4191 preferred_for_speed attribute.
4193 2018-03-14 Richard Biener <rguenther@suse.de>
4195 PR tree-optimization/84830
4196 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4197 with the old one to avoid oscillations.
4199 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4201 PR target/83712
4202 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4203 pseudos.
4204 (assign_by_spills): Return a flag of reload assignment failure.
4205 Do not process the reload assignment failures. Do not spill other
4206 reload pseudos if they has the same reg class. Update n if
4207 necessary.
4208 (lra_assign): Add a return arg. Set up from the result of
4209 assign_by_spills call.
4210 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4211 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4212 usage_insns if it is not NULL.
4213 (spill_hard_reg_in_range): New function.
4214 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4215 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4216 function prototypes.
4217 (lra_assign): Change prototype.
4218 * lra.c (lra): Add code to deal with fails by splitting hard reg
4219 live ranges.
4221 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4223 * config/riscv/riscv.opt (mrelax): New option.
4224 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4225 "norelax" when riscv_mrelax is disabled.
4226 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4228 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4230 PR target/84743
4231 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4232 reassociation for int modes.
4234 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4236 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4237 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4238 for big-endian.
4239 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4240 * config/aarch64/aarch64-sve.md
4241 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4242 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4243 (*extend<mode><Vwide>2): Rename to...
4244 (aarch64_sve_extend<mode><Vwide>2): ...this.
4245 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4246 renaming the old pattern to...
4247 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4248 unsigned packs.
4249 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4250 define_expand, renaming the old pattern to...
4251 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4252 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4253 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4254 account when deciding which SVE instruction the optab should use.
4255 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4257 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4259 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4260 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4261 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4262 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4263 (tlsdesc_small_<mode>): Turn a define_expand and use
4264 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4265 (tlsdesc_small_advsimd_<mode>): ...this.
4266 (tlsdesc_small_sve_<mode>): New pattern.
4268 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4270 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4271 (UNSPEC_UMUL_HIGHPART): New constants.
4272 (MUL_HIGHPART): New int iteraor.
4273 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4274 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4275 define_expand.
4276 (*<su>mul<mode>3_highpart): New define_insn.
4278 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4280 PR lto/84805
4281 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4282 incomplete types.
4284 2018-03-13 Martin Liska <mliska@suse.cz>
4286 PR ipa/84658.
4287 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4288 vector.
4289 (sem_item_optimizer::~sem_item_optimizer): Release it.
4290 (sem_item_optimizer::merge_classes): Register variable aliases.
4291 (sem_item_optimizer::fixup_pt_set): New function.
4292 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4293 * ipa-icf.h: Declare new variables and functions.
4295 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4297 PR middle-end/84834
4298 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4299 integer_pow2p@2 and test integer_pow2p in condition.
4300 (A < 0 ? C : 0): Similarly for @1.
4302 PR middle-end/84831
4303 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4304 characters starting at p contain '\0' character, don't look beyond
4305 that.
4307 PR target/84827
4308 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4309 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4311 PR target/84828
4312 * reg-stack.c (change_stack): Change update_end var from int to
4313 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4314 also call set_block_for_insn on the newly added insns and rescan.
4316 PR target/84786
4317 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4318 on the last operand.
4320 PR c++/84704
4321 * tree.c (stabilize_reference_1): Return save_expr (e) for
4322 STATEMENT_LIST even if it doesn't have side-effects.
4324 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4326 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4328 2018-03-12 Renlin Li <renlin.li@arm.com>
4330 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4331 aarch64_output_scalar_simd_mov_immediate.
4333 2018-03-12 Martin Sebor <msebor@redhat.com>
4335 PR tree-optimization/83456
4336 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4337 for perfectly overlapping calls to memcpy.
4338 (gimple_fold_builtin_memory_chk): Same.
4339 (gimple_fold_builtin_strcpy): Handle no-warning.
4340 (gimple_fold_builtin_stxcpy_chk): Same.
4341 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4343 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4345 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4346 parameter. Use it for SFmode.
4347 (rs6000_function_arg_advance_1): Adjust.
4348 (rs6000_function_arg): Adjust.
4349 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4351 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4353 PR rtl-optimization/84169
4354 PR rtl-optimization/84780
4355 * combine.c (can_combine_p): Check for a 2-insn combination whether
4356 the destination register is used between the two insns, too.
4358 2018-03-12 Richard Biener <rguenther@suse.de>
4360 PR tree-optimization/84803
4361 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4362 for refs DR analysis didn't process.
4364 2018-03-12 Richard Biener <rguenther@suse.de>
4366 PR tree-optimization/84777
4367 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4368 force-vectorize loops ignore whether we are optimizing for size.
4370 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4372 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4373 (TARGET_MD_ASM_ADJUST): Define.
4375 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4376 Kito Cheng <kito.cheng@gmail.com>
4377 Chung-Ju Wu <jasonwucj@gmail.com>
4379 * config/nds32/nds32.c (nds32_compute_stack_frame,
4380 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4381 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4382 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4383 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4384 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4385 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4386 * config/nds32/nds32.md (prologue, epilogue): Use macro
4387 NDS32_V3PUSH_AVAILABLE_P to do checking.
4389 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4391 PR debug/58150
4392 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4393 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4394 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4395 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4396 addition of most attributes on !orig_type_die or the attribute not
4397 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4399 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4400 Chung-Ju Wu <jasonwucj@gmail.com>
4402 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4403 __NDS32_VH__ macro.
4404 * config/nds32/nds32.opt (mvh): New option.
4406 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4407 Chung-Ju Wu <jasonwucj@gmail.com>
4409 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4410 function.
4411 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4412 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4413 definition.
4415 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4416 Chung-Ju Wu <jasonwucj@gmail.com>
4418 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4419 function.
4420 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4421 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4423 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4424 Kito Cheng <kito.cheng@gmail.com>
4425 Chung-Ju Wu <jasonwucj@gmail.com>
4427 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4428 UNSPEC_FFMISM and UNSPEC_FLMISM.
4429 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4430 for ffb, ffmism and flmism.
4431 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4432 (unspec_ffmism): Ditto.
4433 (unspec_flmism): Ditto.
4434 (nds32_expand_builtin_impl): Check if string extension is available.
4435 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4436 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4438 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4440 Reverting patch:
4441 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4443 PR target/83712
4444 * lra-assigns.c (assign_by_spills): Return a flag of reload
4445 assignment failure. Do not process the reload assignment
4446 failures. Do not spill other reload pseudos if they has the same
4447 reg class.
4448 (lra_assign): Add a return arg. Set up from the result of
4449 assign_by_spills call.
4450 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4451 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4452 usage_insns if it is not NULL.
4453 (spill_hard_reg_in_range): New function.
4454 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4455 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4456 function prototypes.
4457 (lra_assign): Change prototype.
4458 * lra.c (lra): Add code to deal with fails by splitting hard reg
4459 live ranges.
4461 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4463 PR target/84807
4464 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4466 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4468 PR debug/84620
4469 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4470 (dw_val_node): Add val_symbolic_view.
4471 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4472 (symview_upper_bound): New.
4473 (new_line_info_table): Initialize symviews_since_reset.
4474 (dwarf2out_source_line): Count symviews_since_reset and set
4475 symview_upper_bound.
4476 (dw_val_equal_p): Handle symview.
4477 (add_AT_symview): New.
4478 (print_dw_val): Handle symview.
4479 (attr_checksum, attr_checksum_ordered): Likewise.
4480 (same_dw_val_p, size_of_die): Likewise.
4481 (value_format, output_die): Likewise.
4482 (add_high_low_attributes): Use add_AT_symview for entry_view.
4483 (dwarf2out_finish): Reset symview_upper_bound, clear
4484 zero_view_p.
4486 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4488 PR target/83969
4489 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4490 Add strict argument and use it.
4491 (rs6000_split_multireg_move): Update for new strict argument.
4492 (mem_operand_gpr): Disallow all non-offsettable addresses.
4493 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4495 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4497 PR target/84772
4498 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4499 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4500 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4502 PR c++/84767
4503 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4504 decl, use remap_type if we want to use the type.
4506 2018-03-09 Martin Sebor <msebor@redhat.com>
4508 PR tree-optimization/84526
4509 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4510 Remove dead code.
4511 (builtin_access::generic_overlap): Be prepared to handle non-array
4512 base objects.
4514 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4516 PR rtl-optimization/84682
4517 * lra-constraints.c (process_address_1): Check is_address flag
4518 for address constraints.
4519 (process_alt_operands): Likewise.
4520 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4521 preprocess_constraints.
4522 * recog.h (preprocess_constraints): Add oploc parameter.
4523 Adjust callers.
4524 * recog.c (preprocess_constraints): Test address_operand for
4525 CT_ADDRESS constraints.
4527 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4529 PR target/83712
4530 * lra-assigns.c (assign_by_spills): Return a flag of reload
4531 assignment failure. Do not process the reload assignment
4532 failures. Do not spill other reload pseudos if they has the same
4533 reg class.
4534 (lra_assign): Add a return arg. Set up from the result of
4535 assign_by_spills call.
4536 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4537 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4538 usage_insns if it is not NULL.
4539 (spill_hard_reg_in_range): New function.
4540 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4541 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4542 function prototypes.
4543 (lra_assign): Change prototype.
4544 * lra.c (lra): Add code to deal with fails by splitting hard reg
4545 live ranges.
4547 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4549 PR target/83193
4550 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4551 Accept complain bool parameter. Only emit errors if it is true.
4552 (arm_parse_cpu_option_name): Likewise.
4553 (arm_target_thumb_only): Adjust callers of the above.
4554 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4555 prototype to take a default true bool parameter.
4556 (arm_parse_arch_option_name): Likewise.
4558 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4559 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4561 PR jit/64089
4562 PR jit/84288
4563 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4564 * configure: Regenerate.
4565 * configure.ac ("linker --version-script option"): New.
4566 ("linker soname option"): New.
4568 2018-03-09 Richard Biener <rguenther@suse.de>
4570 PR tree-optimization/84775
4571 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4572 immediate uses of predicate stmts and mark them modified.
4574 Revert
4575 PR tree-optimization/84178
4576 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4577 to caller.
4578 (version_loop_for_if_conversion): Delay update_ssa call.
4579 (tree_if_conversion): Delay update_ssa until after predicate
4580 insertion.
4582 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4584 PR target/84763
4585 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4586 when the function accesses prior frames.
4588 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4590 PR debug/84456
4591 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4592 gen_llsym, otherwise call maybe_gen_llsym.
4594 PR inline-asm/84742
4595 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4596 has ',' character inside of it.
4598 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4600 PR target/84748
4601 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4602 as clobbering CC_REGNUM.
4604 2018-03-08 Richard Biener <rguenther@suse.de>
4606 PR middle-end/84552
4607 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4608 (follow_copies_to_constant): Do not follow SSA names registered
4609 for update.
4611 2018-03-08 Richard Biener <rguenther@suse.de>
4613 PR tree-optimization/84178
4614 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4615 to caller.
4616 (version_loop_for_if_conversion): Delay update_ssa call.
4617 (tree_if_conversion): Delay update_ssa until after predicate
4618 insertion.
4620 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4622 PR tree-optimization/84178
4623 * tree-if-conv.c (release_bb_predicate): Remove the
4624 the assertion that the stmts have NULL use_ops.
4625 Discard the statements, asserting that they haven't
4626 yet been added to a BB.
4628 2018-03-08 Richard Biener <rguenther@suse.de>
4630 PR tree-optimization/84746
4631 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4632 (phi_translate): Pass in destination ANTIC_OUT set.
4633 (phi_translate_1): Likewise. For a simplified result lookup
4634 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4635 (phi_translate_set): Adjust.
4636 (do_pre_regular_insertion): Likewise.
4637 (do_pre_partial_partial_insertion): Likewise.
4639 2018-03-08 Martin Liska <mliska@suse.cz>
4641 PR gcov-profile/84735
4642 * doc/gcov.texi: Document usage of profile files.
4643 * gcov-io.h: Document changes in the format.
4645 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4647 PR debug/84404
4648 PR debug/84408
4649 * dwarf2out.c (struct dw_line_info_table): Update comments for
4650 view == -1.
4651 (FORCE_RESET_NEXT_VIEW): New.
4652 (FORCE_RESETTING_VIEW_P): New.
4653 (RESETTING_VIEW_P): Check for -1 too.
4654 (ZERO_VIEW_P): Likewise.
4655 (new_line_info_table): Force-reset next view.
4656 (dwarf2out_begin_function): Likewise.
4657 (dwarf2out_source_line): Simplify zero_view_p initialization.
4658 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4659 view directly. Omit view when omitting .loc at line 0.
4661 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4663 PR tree-optimization/84740
4664 * tree-switch-conversion.c (process_switch): Call build_constructors
4665 only if info.phi_count is non-zero.
4667 PR tree-optimization/84739
4668 * tree-tailcall.c (find_tail_calls): Check call arguments against
4669 DECL_ARGUMENTS (current_function_decl) rather than
4670 DECL_ARGUMENTS (func) when checking for tail recursion.
4672 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4674 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4675 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4676 Volker Reichelt's entry and add entries for people that perform
4677 GCC fuzzy testing and report numerous bugs.
4679 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4681 PR target/82411
4682 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4683 readonly data in sdata, if that is disabled.
4684 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4685 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4686 -mreadonly-in-sdata option.
4688 2018-03-07 Martin Sebor <msebor@redhat.com>
4690 PR tree-optimization/84468
4691 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4692 basic block when looking for nul assignment.
4694 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4696 PR target/84277
4697 * except.h (output_function_exception_table): Adjust prototype.
4698 * except.c (output_function_exception_table): Remove FNNAME parameter
4699 and add SECTION parameter. Ouput one part of the table at a time.
4700 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4701 the first part of the exception table and emit unwind directives.
4702 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4703 (i386_pe_seh_cold_init): Likewise.
4704 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4705 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4706 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4707 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4708 * config/i386/winnt.c: Include except.h.
4709 (struct seh_frame_state): Add reg_offset, after_prologue and
4710 in_cold_section fields.
4711 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4712 (i386_pe_seh_cold_init): New function.
4713 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4714 to seh->in_cold_section.
4715 (seh_emit_push): Record the offset of the push.
4716 (seh_emit_save): Record the offet of the save.
4717 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4718 Test seh->after_prologue to disregard the epilogue.
4719 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4720 (i386_pe_end_cold_function): New function.
4722 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4724 PR fortran/84565
4725 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4726 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4728 PR c++/84704
4729 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4730 on tmp_var.
4731 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4732 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4734 PR middle-end/84723
4735 * multiple_target.c: Include tree-inline.h and intl.h.
4736 (expand_target_clones): Diagnose and fail if node->definition and
4737 !tree_versionable_function_p (node->decl).
4739 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4741 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4742 sprint_ul.
4743 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4744 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4745 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4747 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4749 PR target/84710
4750 * combine.c (try_combine): Use reg_or_subregno instead of handling
4751 just paradoxical SUBREGs and REGs.
4753 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4755 * config/arc/arc.c (arc_finalize_pic): Remove function.
4756 (arc_must_save_register): We use single base PIC register, remove
4757 checks to save/restore the PIC register.
4758 (arc_expand_prologue): Likewise.
4759 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4760 Remove.
4761 (arc_verify_short): Likewise.
4762 (arc_attr_type): Likewise.
4763 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4764 (walk_stores): Likewise.
4765 (arc_address_cost): Make it static.
4766 (arc_verify_short): Likewise.
4767 (branch_dest): Likewise.
4768 (arc_attr_type): Likewise.
4769 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4770 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4771 (arc_final_prescan_insn): Remove inserting the nops due to
4772 hardware hazards. It is done in reorg step.
4773 (insn_length_variant_t): Remove.
4774 (insn_length_parameters_t): Likewise.
4775 (arc_insn_length_parameters): Likewise.
4776 (arc_get_insn_variants): Likewise.
4777 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4779 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4781 PR inline-asm/84683
4782 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4783 assertion failure.
4785 PR tree-optimization/84687
4786 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4787 on new_node->decl.
4788 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4790 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4792 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4793 Rename to ppc_speculation_barrier.
4794 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4795 __builtin_ppc_speculation_barrier.
4797 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4799 PR target/84700
4800 * combine.c (combine_simplify_rtx): Don't try to simplify if
4801 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4802 are equal to x.
4804 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4806 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4807 to 32 bytes when compiling for POWER9.
4809 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4811 PR target/84564
4812 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4813 regparm >= 3 with no arg reg available also for calls with
4814 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4816 PR target/84524
4817 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4818 orig,vex.
4819 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4821 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4823 PR target/84264
4824 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4826 2018-03-05 Richard Biener <rguenther@suse.de>
4828 PR tree-optimization/84486
4829 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4830 When inserting a __builtin_assume_aligned call set the LHS
4831 SSA name alignment info accordingly.
4833 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4835 PR tree-optimization/84114
4836 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4837 Avoid reassociation of FLOAT_MODE addition.
4839 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4841 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4842 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4843 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4844 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4845 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4846 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4847 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4848 and -mwbnoinvd.
4849 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4850 __builtin_ia32_wbinvd): New builtins.
4851 (SPECIAL_ARGS2): New.
4852 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4853 (SPECIAL_ARGS2): New.
4854 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4855 (ix86_valid_target_attribute_inner_p): Ditto.
4856 (ix86_init_mmx_sse_builtins): Add special_args2.
4857 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4858 TARGET_WBNOINVD_P): New.
4859 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4860 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4861 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4862 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4863 * config/i386/pconfigintrin.h: New file.
4864 * config/i386/wbnoinvdintrin.h: Ditto.
4865 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4866 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4868 2018-03-05 Richard Biener <rguenther@suse.de>
4870 PR tree-optimization/84670
4871 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4872 member.
4873 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4874 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4875 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4876 doesn't grow if all successors (recursively) were visited at least
4877 once.
4879 2018-03-05 Richard Biener <rguenther@suse.de>
4881 PR tree-optimization/84650
4882 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4883 if executed in the loop pipeline.
4885 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4887 * doc/configfiles.texi (Configuration Files): Move info about
4888 conditionalizing $target-protos.h to...
4889 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4890 differs from $target-protos.h.
4892 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4893 Chung-Ju Wu <jasonwucj@gmail.com>
4895 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4896 * config/nds32/nds32-multiple.md (setmemsi): Define.
4897 * config/nds32/nds32-memory-manipulation.c
4898 (nds32_gen_dup_4_byte_to_word_value): New.
4899 (emit_setmem_word_loop): New.
4900 (emit_setmem_byte_loop): New.
4901 (nds32_expand_setmem_loop): New.
4902 (nds32_expand_setmem_loop_v3m): New.
4903 (nds32_expand_setmem_unroll): New.
4904 (nds32_expand_setmem): New.
4906 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4907 Chung-Ju Wu <jasonwucj@gmail.com>
4909 * config/nds32/nds32-memory-manipulation.c
4910 (nds32_emit_load_store): New.
4911 (nds32_emit_post_inc_load_store): New.
4912 (nds32_emit_mem_move): New.
4913 (nds32_emit_mem_move_block): New.
4914 (nds32_expand_movmemsi_loop_unknown_size): New.
4915 (nds32_expand_movmemsi_loop_known_size): New.
4916 (nds32_expand_movmemsi_loop): New.
4917 (nds32_expand_movmemsi_unroll): New.
4918 (nds32_expand_movmemqi): Rename ...
4919 (nds32_expand_movmemsi): ... to this.
4920 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4921 (movmemsi): ... to this.
4922 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4923 (nds32_expand_movmemsi): ... to this.
4925 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4926 Monk Chiang <sh.chiang04@gmail.com>
4927 Chung-Ju Wu <jasonwucj@gmail.com>
4929 * config/nds32/nds32-protos.h
4930 (nds32_expand_load_multiple): New arguments.
4931 (nds32_expand_store_multiple): Ditto.
4932 (nds32_valid_multiple_load_store): Rename ...
4933 (nds32_valid_multiple_load_store_p): ... to this.
4934 * config/nds32/nds32-memory-manipulation.c
4935 (nds32_expand_load_multiple): Refine implementation.
4936 (nds32_expand_store_multiple): Ditto.
4937 * config/nds32/nds32-multiple.md
4938 (load_multiple): Update nds32_expand_load_multiple interface.
4939 (store_multiple): Update nds32_expand_store_multiple interface.
4940 * config/nds32/nds32-predicates.c
4941 (nds32_valid_multiple_load_store): Rename ...
4942 (nds32_valid_multiple_load_store_p): ... to this and refine
4943 implementation.
4944 * config/nds32/predicates.md
4945 (nds32_load_multiple_and_update_address_operation): New predicate.
4946 (nds32_store_multiple_and_update_address_operation): New predicate.
4948 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4949 Chung-Ju Wu <jasonwucj@gmail.com>
4951 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4952 (combo): New attribute.
4953 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4955 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4957 * config/nds32/nds32.opt: Change -mcmodel= default value.
4959 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4960 Monk Chiang <sh.chiang04@gmail.com>
4961 Chung-Ju Wu <jasonwucj@gmail.com>
4963 * config/nds32/constants.md (unspec_element): New enum.
4964 * config/nds32/constraints.md (Umw): New constraint.
4965 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4966 * config/nds32/nds32-intrinsic.md: Likewise.
4967 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4968 (nds32_valid_smw_lwm_base_p): New.
4969 (nds32_output_smw_single_word): New.
4970 (nds32_output_lmw_single_word): New.
4971 (nds32_expand_unaligned_load): New.
4972 (nds32_expand_unaligned_store): New.
4973 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4974 (nds32_output_smw_single_word): Declare.
4975 (nds32_output_lmw_single_word): Declare.
4976 (nds32_expand_unaligned_load): Declare.
4977 (nds32_expand_unaligned_store): Declare.
4978 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4979 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4980 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4981 NDS32_BUILTIN_UASTORE_DW.
4982 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4983 predicate.
4985 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4986 Kito Cheng <kito.cheng@gmail.com>
4987 Chung-Ju Wu <jasonwucj@gmail.com>
4989 * config/nds32/nds32-intrinsic.c
4990 (nds32_expand_builtin_null_ftype_reg): Delete.
4991 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4992 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4993 (nds32_read_argument): New.
4994 (nds32_legitimize_target): Ditto.
4995 (nds32_legitimize_argument): Ditto.
4996 (nds32_check_constant_argument): Ditto.
4997 (nds32_expand_unop_builtin): Ditto.
4998 (nds32_expand_unopimm_builtin): Ditto.
4999 (nds32_expand_binop_builtin): Ditto.
5000 (nds32_builtin_decl_impl): Ditto.
5001 (builtin_description): Ditto.
5002 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5003 (nds32_init_builtins_impl): Ditto.
5004 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5005 (nds32_builtin_decl): New.
5006 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5007 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5009 2018-03-02 Jeff Law <law@redhat.com>
5011 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5012 (redundant_insn, fill_simple_delay_slots): Likewise.
5013 (fill_slots_from_thread): Likewise.
5014 * resource.c (mark_referenced_resources): Likewise.
5015 (mark_set_resources, find_dead_or_set_registers): Likewise.
5017 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5019 * substring-locations.h (format_warning_va): Formatting fix for
5020 ATTRIBUTE_GCC_DIAG.
5021 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5022 argument.
5023 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5024 * substring-locations.c: Include intl.h.
5025 (format_warning_va): Turned into small wrapper around
5026 format_warning_n_va, renamed to ...
5027 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5028 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5029 use ngettext.
5030 (format_warning_at_substring_n): New function.
5031 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5032 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5033 format_warning_at_substring with just a shorter name instead of
5034 const function pointer.
5035 (fmtwarn_n): New function.
5036 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5037 appropriate, get rid of all the fmtstr temporaries, move conditionals
5038 with G_() wrapped string literals directly into fmtwarn arguments,
5039 cast dir.len to (int), formatting fixes.
5041 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5043 * doc/invoke.texi: Remove "Cilk Plus" references.
5045 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5046 Richard Biener <rguenther@suse.de>
5048 PR ipa/84628
5049 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5050 for error or warning attributes if CALL_FROM_THUNK_P is set.
5051 Formatting fixes.
5053 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5055 PR target/56540
5056 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5057 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5059 PR target/56540
5060 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5061 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5063 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5064 instead of -1U in last predictors element's probability member.
5066 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5068 PR ipa/83983
5069 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5070 arguments if they are comparable.
5072 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5074 PR tree-optimization/84634
5075 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5076 masks and masked_loop_p with a single loop_masks, making sure it's
5077 null for bb vectorization.
5079 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5081 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5082 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5083 loop->force_vectorize to check whether there is no alias.
5085 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5087 PR target/84614
5088 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5089 prototypes.
5090 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5091 comments.
5092 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5093 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5094 instead of a loop around prev_real_insn.
5095 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5096 prev_real_insn.
5098 PR inline-asm/84625
5099 * config/i386/i386.c (ix86_print_operand): Use conditional
5100 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5101 zero vector.
5103 2018-03-02 Richard Biener <rguenther@suse.de>
5105 PR tree-optimization/84427
5106 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5107 (bitmap_set_subtract_values): Rewrite to handle multiple
5108 exprs per value.
5109 (clean): Likewise.
5110 (prune_clobbered_mems): Likewise.
5111 (phi_translate): Take edge instead of pred/phiblock.
5112 (phi_translate_1): Likewise.
5113 (phi_translate_set): Likewise. Insert all translated
5114 exprs for a value into the set, keeping possibly multiple
5115 expressions per value.
5116 (compute_antic_aux): Adjust for phi_translate changes.
5117 When intersecting union the expressions and prune those
5118 not in the final value set, keeping possibly multiple
5119 expressions per value. Do not use value-insertion
5120 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5121 all expressions. Add verification that the value-sets
5122 only shrink during iteration.
5123 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5124 (do_pre_regular_insertion): Likewise.
5125 (do_pre_partial_partial_insertion): Likewise.
5127 2018-03-02 Richard Biener <rguenther@suse.de>
5129 PR target/82005
5130 * config/darwin.c (saved_debug_info_level): New static global.
5131 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5132 (darwin_asm_lto_end): Restore debug info generation settings.
5134 2018-03-01 Martin Liska <mliska@suse.cz>
5136 PR sanitizer/82484
5137 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5138 volatile arguments.
5140 2018-03-01 Richard Biener <rguenther@suse.de>
5142 PR debug/84645
5143 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5144 type annotation with LTO when debug was disabled at compile-time.
5146 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5148 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5149 XINT with INTVAL.
5150 (mips_final_postscan_insn): Likewise.
5152 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5154 PR rtl-optimization/84528
5155 * alias.c (init_alias_target): Add commentary.
5156 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5157 a unique base value if the frame pointer is not eliminated
5158 to the stack pointer.
5160 2018-03-01 Tom de Vries <tom@codesourcery.com>
5162 PR rtl-optimization/83327
5163 * lra-int.h (hard_regs_spilled_into): Declare.
5164 * lra.c (hard_regs_spilled_into): Define.
5165 (init_reg_info): Init hard_regs_spilled_into.
5166 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5167 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5168 (process_bb_lives): Handle hard_regs_spilled_into.
5169 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5170 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5172 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5174 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5175 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5176 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5177 * config/rs6000/aix72.h: New file.
5179 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5181 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5182 instead of warning_at with conditional singular and plural messages
5183 where possible.
5185 PR target/52991
5186 * stor-layout.c (update_alignment_for_field): For
5187 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5188 && !DECL_PACKED (field), do the alignment update, just use
5189 only desired_align instead of MAX (type_align, desired_align)
5190 as the alignment.
5191 (place_field): Don't do known_align < desired_align handling
5192 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5193 is non-NULL, instead do it after rli->prev_field handling and
5194 only if not within a bitfield word. For DECL_PACKED (field)
5195 use type_align of BITS_PER_UNIT.
5197 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5199 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5200 superfluous parentheses and trailing spaces.
5202 2018-02-28 Richard Biener <rguenther@suse.de>
5204 PR tree-optimization/84584
5205 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5206 SCoPs with fake exit edge.
5208 2018-02-28 Martin Liska <mliska@suse.cz>
5210 PR testsuite/84597
5211 * timevar.c (timer::print): Fix format to properly print 100%
5212 values.
5214 2018-02-28 Richard Biener <rguenther@suse.de>
5216 PR middle-end/84607
5217 * genmatch.c (capture_info::walk_match): Do not mark
5218 captured expressions without operands as expr_p given
5219 they act more like predicates and should be subject to
5220 "lost tail" side-effect preserving.
5222 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5224 PR rtl-optimization/81611
5225 * auto-inc-dec.c (attempt_change): Move dead note from
5226 mem_insn if it's the next use of regno
5227 (find_address): Take address use of reg holding
5228 non-incremented value. Add parm to limit search to the named
5229 reg only.
5230 (merge_in_block): Attempt to use a mem insn that is the next
5231 use of the original regno.
5233 2018-02-27 Martin Sebor <msebor@redhat.com>
5235 PR c++/83871
5236 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5237 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5239 2018-02-27 Martin Sebor <msebor@redhat.com>
5241 PR translation/84207
5242 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5243 n argument to unsigned HOST_WIDE_INT.
5244 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5245 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5246 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5247 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5249 2018-02-27 Richard Biener <rguenther@suse.de>
5251 PR tree-optimization/84512
5252 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5253 Do not use the estimate returned from record_stmt_cost for
5254 the scalar iteration cost but sum properly using add_stmt_cost.
5256 2018-02-27 Richard Biener <rguenther@suse.de>
5258 PR tree-optimization/84466
5259 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5260 Adjust last change to less strictly validate use operands.
5262 2018-02-27 Martin Liska <mliska@suse.cz>
5264 PR gcov-profile/84548
5265 * gcov.c (process_file): Allow partial overlap and consider it
5266 also as group functions.
5267 (output_lines): Properly calculate range of lines for a group.
5269 2018-02-27 Martin Liska <mliska@suse.cz>
5271 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5272 'ggc' suffixes. Change first column width.
5273 (timer::print): Fix formatting of the column.
5275 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5277 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5278 preserve inline entry blocks for the sake of debug inline
5279 entry point markers alone.
5280 (remove_unused_locals): Suggest in comments a better place to
5281 force the preservation of inline entry blocks that are
5282 otherwise unused, but do not preserve them.
5284 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5286 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5288 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5290 PR target/84039
5291 * config/i386/constraints.md (Bs): Replace
5292 ix86_indirect_branch_register with
5293 TARGET_INDIRECT_BRANCH_REGISTER.
5294 (Bw): Likewise.
5295 * config/i386/i386.md (indirect_jump): Likewise.
5296 (tablejump): Likewise.
5297 (*sibcall_memory): Likewise.
5298 (*sibcall_value_memory): Likewise.
5299 Peepholes of indirect call and jump via memory: Likewise.
5300 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5301 (*sibcall_value_GOT_32): Likewise.
5302 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5303 (GOT_memory_operand): Likewise.
5304 (call_insn_operand): Likewise.
5305 (sibcall_insn_operand): Likewise.
5306 (GOT32_symbol_operand): Likewise.
5307 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5309 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5311 PR rtl-optimization/83496
5312 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5313 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5314 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5315 redundant insn, if any.
5316 (relax_delay_slots): Likewise.
5317 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5319 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5321 PR tree-optimization/83965
5322 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5323 that grouped statements are part of a reduction chain. Return
5324 true if the statement is not marked as a reduction itself but
5325 is part of a group.
5326 (vect_recog_dot_prod_pattern): Don't check whether the statement
5327 is part of a group here.
5328 (vect_recog_sad_pattern): Likewise.
5329 (vect_recog_widen_sum_pattern): Likewise.
5331 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5333 PR debug/84545
5334 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5336 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5338 PR target/84530
5339 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5340 the bool argument.
5341 (ix86_output_indirect_function_return): New prototype.
5342 (ix86_split_simple_return_pop_internal): Likewise.
5343 * config/i386/i386.c (indirect_return_via_cx): New.
5344 (indirect_return_via_cx_bnd): Likewise.
5345 (indirect_thunk_name): Handle return va CX_REG.
5346 (output_indirect_thunk_function): Create alias for
5347 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5348 (ix86_output_indirect_jmp): Remove the bool argument.
5349 (ix86_output_indirect_function_return): New function.
5350 (ix86_split_simple_return_pop_internal): Likewise.
5351 * config/i386/i386.md (*indirect_jump): Don't pass false
5352 to ix86_output_indirect_jmp.
5353 (*tablejump_1): Likewise.
5354 (simple_return_pop_internal): Change it to define_insn_and_split.
5355 Call ix86_split_simple_return_pop_internal to split it for
5356 -mfunction-return=.
5357 (simple_return_indirect_internal): Call
5358 ix86_output_indirect_function_return instead of
5359 ix86_output_indirect_jmp.
5361 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5363 PR bootstrap/84405
5364 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5365 memset and value initialization afterwards.
5367 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5369 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5371 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5373 PR target/84521
5374 * common/config/aarch64/aarch64-common.c
5375 (aarch_option_optimization_table[]): Switch
5376 off fomit-frame-pointer
5378 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5379 Chung-Ju Wu <jasonwucj@gmail.com>
5381 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5382 volatile memory.
5383 (store_multiple): Ditto.
5385 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5387 * config.gcc: Add --with-cpu support for nds32 target.
5388 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5389 * config/nds32/nds32.opt: Add -mcpu= option.
5391 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5393 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5394 isel=yes): Warn for these deprecated options.
5396 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5398 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5399 ISA_2_5_MASKS_EMBEDDED.
5401 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5403 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5404 p->max as pointers rather than using iterative_hash_expr.
5406 2018-02-23 Carl Love <cel@us.ibm.com>
5408 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5409 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5410 BU_P8V_OVERLOAD_2.
5411 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5412 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5413 P8V_BUILTIN_VEC_VUNSIGNED2.
5415 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5417 PR target/81572
5418 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5419 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5420 LRA_UNKNOWN_ALT.
5421 * lra-constraints.c (curr_insn_transform): Set up
5422 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5423 LRA_UNKNOWN_ALT.
5424 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5425 * lra-eliminations.c (spill_pseudos): Ditto.
5426 (process_insn_for_elimination): Ditto.
5427 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5428 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5429 LRA_NON_CLOBBERED_ALT.
5431 2018-02-22 Martin Sebor <msebor@redhat.com>
5433 PR tree-optimization/84480
5434 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5435 to maybe_diag_stxncpy_trunc. Call it.
5436 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5437 from gimple_fold_builtin_strcpy. Print inlining stack.
5438 (handle_builtin_stxncpy): Print inlining stack.
5439 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5441 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5443 PR target/84176
5444 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5445 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5446 and -fcheck-pointer-bounds are used together.
5447 (indirect_thunk_prefix): New enum.
5448 (indirect_thunk_need_prefix): New function.
5449 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5450 "_nt" instead of "_bnd" for NOTRACK prefix.
5451 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5452 (output_indirect_thunk_function): Likewise.
5453 (): Likewise.
5454 (ix86_code_end): Update output_indirect_thunk_function calls.
5455 (ix86_output_indirect_branch_via_reg): Replace
5456 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5457 (ix86_output_indirect_branch_via_push): Likewise.
5458 (ix86_output_function_return): Likewise.
5459 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5460 incompatible with -fcf-protection=branch and
5461 -fcheck-pointer-bounds.
5463 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5465 PR target/83335
5466 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5467 Change gcc_assert call to output_operand_lossage.
5469 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5471 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5473 2018-02-22 DJ Delorie <dj@redhat.com>
5474 Sebastian Perta <sebastian.perta@renesas.com>
5475 Oleg Endo <olegendo@gcc.gnu.org>
5477 * config/rx/rx.c (rx_rtx_costs): New function.
5478 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5480 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5482 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5484 2018-02-22 Martin Liska <mliska@suse.cz>
5486 PR driver/83193
5487 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5488 Add "native" as a possible value.
5490 2018-02-22 Martin Liska <mliska@suse.cz>
5492 PR driver/83193
5493 * config/i386/i386.c (ix86_option_override_internal):
5494 Add "native" as a possible value for -march and -mtune.
5496 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5498 PR target/84502
5499 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5500 to all type variants.
5502 PR tree-optimization/84503
5503 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5504 width as info->bitpos + info->bitsize - start.
5505 (merged_store_group::merge_overlapping): Simplify width computation.
5506 (check_no_overlap): New function.
5507 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5508 start + width and last_order of the group, fail if check_no_overlap
5509 fails.
5510 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5511 to group if check_no_overlap fails.
5513 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5515 * config/rs6000/altivec.md: Delete contraint arguments to
5516 define_expand, define_split, and define_peephole2, and in
5517 define_insn_and_split if always unused.
5518 * config/rs6000/darwin.md: Ditto.
5519 * config/rs6000/dfp.md: Ditto.
5520 * config/rs6000/rs6000.md: Ditto.
5521 * config/rs6000/sync.md: Ditto.
5522 * config/rs6000/vector.md: Ditto.
5523 * config/rs6000/vsx.md: Ditto.
5525 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5527 * config/rs6000/altivec.md: Write output control strings as braced
5528 blocks instead of double-quoted strings.
5529 * config/rs6000/darwin.md: Ditto.
5530 * config/rs6000/rs6000.md: Ditto.
5531 * config/rs6000/vector.md: Ditto.
5532 * config/rs6000/vsx.md: Ditto.
5534 2018-02-21 Jason Merrill <jason@redhat.com>
5536 PR c++/84314 - ICE with templates and fastcall attribute.
5537 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5539 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5541 * ipa-cp.c (determine_versionability): Fix comment typos.
5543 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5545 PR c/84229
5546 * ipa-cp.c (determine_versionability): Do not version functions caling
5547 va_arg_pack.
5549 2018-02-21 Martin Liska <mliska@suse.cz>
5551 PR driver/83193
5552 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5553 Add "native" as a possible value.
5554 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5555 the macro when native cpu detection is available.
5557 2018-02-21 Martin Liska <mliska@suse.cz>
5559 PR driver/83193
5560 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5561 Add "native" as a possible value.
5562 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5563 when native cpu detection is available.
5565 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5566 Martin Sebor <msebor@redhat.com>
5568 PR tree-optimization/84478
5569 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5570 false.
5571 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5572 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5573 support which is conservatively correct, for 2 only stay conservative
5574 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5575 argument to the 2 argument get_range_strlen, adjust 6 arg
5576 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5577 false.
5578 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5579 (gimple_fold_builtin_strlen): Pass true as last argument to
5580 get_range_strlen.
5582 2018-02-20 Martin Sebor <msebor@redhat.com>
5584 PR middle-end/84095
5585 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5586 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5587 (builtin_memref::builtin_memref): Factor out parts into
5588 set_base_and_offset and call it.
5590 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5592 PR middle-end/84406
5593 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5594 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5595 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5596 search at the associated MODE_INT.
5598 2018-02-20 Jeff Law <law@redhat.com>
5600 PR middle-end/82123
5601 PR tree-optimization/81592
5602 PR middle-end/79257
5603 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5604 for range data rather than using global data.
5605 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5606 range data rather than using global data.
5607 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5608 pass it to children as needed.
5609 (struct directive::fmtresult): Similarly.
5610 (struct directive::set_width): Similarly.
5611 (struct directive::set_precision): Similarly.
5612 (format_integer, format_directive, parse_directive): Similarly.
5613 (format_none): Accept unnamed vr_values parameter.
5614 (format_percent, format_floating, format_character): Similarly.
5615 (format_string, format_plain): Similarly.
5616 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5617 the EVRP range analyzer for range data rather than using global data.
5618 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5619 gimple-ssa-evrp-analyze.h
5620 (class sprintf_dom_walker): Add after_dom_children member function.
5621 Add evrp_range_analyzer member.
5622 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5623 range analyzer as needed.
5624 (sprintf_dom_walker::after_dom_children): New member function.
5625 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5626 if not optimizing.
5627 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5628 (evrp_range_analyzer::pop_to_marker): Likewise.
5630 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5632 PR tree-optimization/84419
5633 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5634 with the required type if its current type is compatible but
5635 different.
5637 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5639 PR middle-end/82004
5640 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5641 after vectorization.
5643 2018-02-20 Martin Liska <mliska@suse.cz>
5645 PR driver/83193
5646 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5647 possible values if we don't have a hint.
5649 2018-02-20 Martin Liska <mliska@suse.cz>
5651 PR c/84310
5652 PR target/79747
5653 * final.c (shorten_branches): Build align_tab array with one
5654 more element.
5655 * opts.c (finish_options): Add alignment option limit check.
5656 (MAX_CODE_ALIGN): Likewise.
5657 (MAX_CODE_ALIGN_VALUE): Likewise.
5658 * doc/invoke.texi: Document maximum allowed option value for
5659 all -falign-* options.
5661 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5663 PR target/84146
5664 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5665 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5666 * var-tracking.c (emit_note_insn_var_location): Remove all references
5667 to NOTE_INSN_CALL_ARG_LOCATION.
5668 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5669 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5670 Use copy_rtx_if_shared.
5671 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5672 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5673 (dwarf2out_var_location): Remove handling of
5674 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5675 on call_insn.
5676 * final.c (final_scan_insn): Remove all references to
5677 NOTE_INSN_CALL_ARG_LOCATION.
5678 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5679 before dumping final insns.
5680 * except.c (emit_note_eh_region_end): Remove all references to
5681 NOTE_INSN_CALL_ARG_LOCATION.
5682 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5683 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5684 * config/arc/arc.c (hwloop_optimize): Likewise.
5685 * config/arm/arm.c (create_fix_barrier): Likewise.
5686 * config/s390/s390.c (s390_chunkify_start): Likewise.
5687 * config/sh/sh.c (find_barrier): Likewise.
5688 * config/i386/i386.c (rest_of_insert_endbranch,
5689 ix86_seh_fixup_eh_fallthru): Likewise.
5690 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5691 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5692 * config/frv/frv.c (frv_function_prologue): Likewise.
5693 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5694 reg note.
5695 (note_outside_basic_block_p): Remove all references to
5696 NOTE_INSN_CALL_ARG_LOCATION.
5697 * gengtype.c (adjust_field_rtx_def): Likewise.
5698 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5699 Likewise.
5700 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5701 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5703 PR c++/84444
5704 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5705 is ADDR_EXPR.
5707 PR tree-optimization/84452
5708 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5709 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5710 is NULL.
5712 2018-02-19 Martin Liska <mliska@suse.cz>
5714 PR sanitizer/82183
5715 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5717 2018-02-19 Martin Liska <mliska@suse.cz>
5718 Richard Sandiford <richard.sandiford@linaro.org>
5720 PR tree-optimization/82491
5721 * gimple-fold.c (get_base_constructor): Make earlier bail out
5722 to prevent ubsan.
5724 2018-02-19 Carl Love <cel@us.ibm.com>
5726 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5727 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5728 BU_P8V_OVERLOAD_1.
5729 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5730 P8V_BUILTIN_VEC_NEG.
5732 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5734 * config/rl78/rl78.md (movdf): New define expand.
5736 2018-02-19 Martin Liska <mliska@suse.cz>
5738 PR other/80589
5739 * doc/invoke.texi: Fix typo.
5740 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5742 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5744 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5745 handle rs6000_single_float and rs6000_double_float specially for
5746 e500 family CPUs.
5748 2018-02-16 Jeff Law <law@redhat.com>
5750 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5751 (pop_regs): Use it.
5753 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5755 PR ipa/84425
5756 * ipa-inline.c (inline_small_functions): Fix a typo.
5758 2018-02-16 Nathan Sidwell <nathan@acm.org>
5760 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5762 2018-02-16 Carl Love <cel@us.ibm.com>
5764 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5765 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5766 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5767 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5768 expansion to P8V_BUILTIN_VEC_FLOAT2.
5770 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5772 PR rtl-optimization/70023
5773 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5774 src_regno into account.
5776 2018-02-16 Carl Love <cel@us.ibm.com>
5778 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5779 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5780 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5781 * config/rs6000/rs6000.c: Remove case statements for
5782 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5783 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5784 and P9V_BUILTIN_VEC_VINSERT4B.
5785 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5786 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5787 * config/rs6000/vsx.md:
5788 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5789 vec_insert4b.
5791 2018-02-16 Carl Love <cel@us.ibm.com>
5793 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5794 vec_insert4b.
5795 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5796 definitions.
5797 * config/rs6000/rs6000-c.c: Add the definitions for
5798 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5799 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5800 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5801 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5802 definition for insert4b and define insn *insert3b_internal.
5803 * doc/extend.texi: Add documentation for vec_extract4b.
5805 2018-02-16 Nathan Sidwell <nathan@acm.org>
5807 * doc/extend.texi (Backwards Compatibility): Mention friend
5808 injection. Note for-scope is deprecated.
5809 * doc/invoke.texi (-ffriend-injection): Deprecate.
5811 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5813 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5814 that moved to I2, also allow destinations that are a paradoxical
5815 subreg (instead of a normal reg).
5817 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5819 PR target/83831
5820 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5821 to QImode.
5823 2018-02-16 Richard Biener <rguenther@suse.de>
5825 PR tree-optimization/84037
5826 PR tree-optimization/84016
5827 PR target/82862
5828 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5829 Adjust vec_construct for the fact we need additional higher latency
5830 128bit inserts for AVX256 and AVX512 vector builds.
5831 (ix86_add_stmt_cost): Scale vector construction cost for
5832 elementwise loads.
5834 2018-02-16 Richard Biener <rguenther@suse.de>
5836 PR tree-optimization/84417
5837 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5838 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5839 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5841 2018-02-16 Martin Liska <mliska@suse.cz>
5843 PR sanitizer/84307
5844 * internal-fn.def (ASAN_CHECK): Set proper flags.
5845 (ASAN_MARK): Likewise.
5847 2018-02-16 Julia Koval <julia.koval@intel.com>
5849 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5850 from PTA_CANNONLAKE.
5852 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5854 PR target/84272
5855 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5856 Use ++iter rather than iter++ for std::list iterators.
5857 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5858 defer deleting them until all nodes in the forest are processed. Do
5859 free even leaf nodes. Change to_process into auto_vec.
5861 PR bootstrap/84405
5862 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5863 * vec.h (vec_default_construct): Use memset instead of placement new
5864 if BROKEN_VALUE_INITIALIZATION is defined.
5865 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5866 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5867 is defined.
5869 PR rtl-optimization/83723
5870 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5871 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5872 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5873 recursive calls.
5874 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5875 callers.
5876 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5878 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5880 PR rtl-optimization/81443
5881 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5882 from inner REGs to paradoxical SUBREGs.
5884 2018-02-16 Richard Biener <rguenther@suse.de>
5886 PR tree-optimization/84399
5887 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5888 For operands we can analyze at their definition make sure we can
5889 analyze them at each use as well.
5891 2018-02-16 Richard Biener <rguenther@suse.de>
5893 PR tree-optimization/84190
5894 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5895 volatile accesses if the decl isn't volatile.
5897 2018-02-15 Jason Merrill <jason@redhat.com>
5899 PR c++/84314 - ICE with templates and fastcall attribute.
5900 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5901 TYPE_CANONICAL on an existing type.
5903 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5905 PR tree-optimization/84383
5906 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5907 dstoff nor call operand_equal_p if dstbase is NULL.
5909 PR tree-optimization/84334
5910 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5911 also a CONSTANT_CLASS_P, punt.
5913 2018-02-14 Jim Wilson <jimw@sifive.com>
5915 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5916 first SMALL_OPERAND check. New local min_second_step. Move assert
5917 to where locals are set. Add TARGET_RVC support.
5918 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5920 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5922 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5924 2018-02-14 Martin Sebor <msebor@redhat.com>
5926 PR tree-optimization/83698
5927 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5928 arrays constrain the offset range to their bounds.
5929 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5930 (builtin_access::overlap): Avoid setting the size of overlap if it's
5931 already been set.
5932 (maybe_diag_overlap): Also consider arrays when deciding what values
5933 of offsets to include in diagnostics.
5935 2018-02-14 Martin Sebor <msebor@redhat.com>
5937 PR c/84108
5938 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5939 that correspond to the kind of a declaration.
5941 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5943 PR target/83984
5944 * config/pa/pa.md: Load address of PIC label using the linkage table
5945 if the label is nonlocal.
5947 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5949 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5950 warning message if user requests -maltivec=be.
5951 * doc/invoke.texi: Document deprecation of -maltivec=be.
5953 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5955 PR target/84220
5956 * config/rs6000/rs6000-c.c: Update definitions for
5957 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5958 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5960 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5962 PR target/84239
5963 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5964 add _get_ssp intrinsics. Remove argument from
5965 __builtin_ia32_rdssp[d|q].
5966 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5967 * config/i386/i386-builtin.def: Remove argument from
5968 __builtin_ia32_rdssp[d|q].
5969 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5970 ix86_expand_special_args_builtin for _rdssp[d|q].
5971 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5972 Clear register before usage.
5973 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5974 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5976 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5978 PR tree-optimization/84357
5979 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5980 operand 1 of an ARRAY_REF too.
5982 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5984 PR target/83831
5985 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5986 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5987 declarations.
5988 (set_of_reg): New struct.
5989 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5990 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5991 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5992 functions.
5993 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5994 Split into bitclr, bitset, bitinvert patterns if appropriate.
5995 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5996 use rx_fuse_in_memory_bitop.
5997 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5998 to named insn, correct maximum insn length.
6000 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6002 PR target/79242
6003 * machmode.def: Define a complex mode for PARTIAL_INT.
6004 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6005 MODE_PARTIAL_INT.
6006 * doc/rtl.texi: Document CSPImode.
6007 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6008 handling.
6009 (msp430_hard_regno_nregs_with_padding): Likewise.
6011 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6013 PR target/84279
6014 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6016 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6018 PR rtl-optimization/84169
6019 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6020 we generated a parallel as new i3 and we split that to new i2 and i3
6021 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6022 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6023 those to i2, not i1. Partially rewrite this scan code.
6025 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6027 PR c/82210
6028 * stor-layout.c (place_field): For variable length fields, adjust
6029 offset_align afterwards not just based on the field's alignment,
6030 but also on the size.
6032 PR middle-end/84309
6033 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6034 of exps and logs in the use_exp2 case.
6036 2018-02-13 Jeff Law <law@redhat.com>
6038 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6039 entry for "vector".
6041 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6042 ARGS as unused.
6044 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6046 PR debug/84342
6047 PR debug/84319
6048 * common.opt (gas-loc-support, gas-locview-support): New.
6049 (ginline-points, ginternal-reset-location-views): New.
6050 * doc/invoke.texi: Document them. Use @itemx where intended.
6051 (gvariable-location-views): Adjust.
6052 * target.def (reset_location_view): New.
6053 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6054 (TARGET_RESET_LOCATION_VIEW): New.
6055 * doc/tm.texi: Rebuilt.
6056 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6057 (dwarf2out_default_as_locview_support): New.
6058 (output_asm_line_debug_info): Use option variables.
6059 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6060 (output_loc_list): Likewise.
6061 (add_high_low_attributes): Check option variables.
6062 Don't output entry view attribute in strict mode.
6063 (gen_inlined_subroutine_die): Check option variables.
6064 (dwarf2out_inline_entry): Likewise.
6065 (init_sections_and_labels): Likewise.
6066 (dwarf2out_early_finish): Likewise.
6067 (maybe_reset_location_view): New, from...
6068 (dwarf2out_var_location): ... here. Call it.
6069 * debug.h (dwarf2out_default_as_loc_support): Declare.
6070 (dwarf2out_default_as_locview_support): Declare.
6071 * hooks.c (hook_int_rtx_insn_0): New.
6072 * hooks.h (hook_int_rtx_insn_0): Declare.
6073 * toplev.c (process_options): Take -gas-loc-support and
6074 -gas-locview-support from dwarf2out. Enable
6075 -gvariable-location-views by default only with locview
6076 assembler support. Enable -ginternal-reset-location-views by
6077 default only if the target defines the corresponding hook.
6078 Enable -ginline-points by default if location views are
6079 enabled; force it disabled if statement frontiers are
6080 disabled.
6081 * tree-inline.c (expand_call_inline): Check option variables.
6082 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6084 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6086 PR tree-optimization/84321
6087 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6088 handling. Also check whether the anti-range contains any values
6089 that satisfy the mask; switch to a VR_RANGE if not.
6091 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6093 PR sanitizer/84340
6094 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6096 2018-02-13 Martin Jambor <mjambor@suse.cz>
6098 PR c++/83990
6099 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6100 of call statements, also set location of a load to a temporary.
6102 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6104 * config/rl78/rl78.c (add_vector_labels): New function.
6105 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6106 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6107 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6108 which checks that no arguments are passed.
6109 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6110 * doc/extend.texi: Documentation for the new attribute.
6112 2018-02-13 Andreas Schwab <schwab@suse.de>
6114 * config/riscv/linux.h (CPP_SPEC): Define.
6116 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6118 PR target/84335
6119 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6120 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6121 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6122 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6123 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6124 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6125 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6126 temporarily for AES and PCLMUL builtins.
6128 PR tree-optimization/84339
6129 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6130 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6131 Formatting fixes.
6133 PR middle-end/84309
6134 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6135 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6136 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6137 inline function.
6138 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6139 inline function.
6140 * omp-simd-clone.h: New file.
6141 * omp-simd-clone.c: Include omp-simd-clone.h.
6142 (expand_simd_clones): No longer static.
6143 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6144 cgraph.h and omp-simd-clone.h.
6145 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6146 (vect_recog_widen_shift_pattern): Formatting fix.
6147 (vect_pattern_recog_1): Don't check optab for calls.
6149 PR target/84336
6150 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6151 operands[2] into a REG before using gen_lowpart on it.
6153 2018-02-12 Jeff Law <law@redhat.com>
6155 PR target/83760
6156 * config/sh/sh.c (find_barrier): Consider a sibling call
6157 a barrier as well.
6159 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6160 successfully back substituting a reg.
6162 2018-02-12 Richard Biener <rguenther@suse.de>
6164 PR tree-optimization/84037
6165 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6166 parameter, move visited init to caller.
6167 (vect_slp_analyze_operations): Separate cost from validity
6168 check, initialize visited once for all instances.
6169 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6170 for all instances.
6171 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6172 out an assert.
6173 (vect_model_promotion_demotion_cost): Likewise.
6174 (vectorizable_bswap): Guard cost modeling with !slp_node
6175 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6176 SLP stmts.
6177 (vectorizable_call): Likewise.
6178 (vectorizable_conversion): Likewise.
6179 (vectorizable_assignment): Likewise.
6180 (vectorizable_shift): Likewise.
6181 (vectorizable_operation): Likewise.
6182 (vectorizable_store): Likewise.
6183 (vectorizable_load): Likewise.
6184 (vectorizable_condition): Likewise.
6185 (vectorizable_comparison): Likewise.
6187 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6189 PR sanitizer/84307
6190 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6191 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6192 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6194 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6196 PR middle-end/83665
6197 * params.def (inline-min-speedup): Increase from 8 to 15.
6198 (max-inline-insns-auto): Decrease from 40 to 30.
6199 * ipa-split.c (consider_split): Add some buffer for function to
6200 be considered inlining candidate.
6201 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6202 default values.
6204 2018-02-12 Richard Biener <rguenther@suse.de>
6206 PR tree-optimization/84037
6207 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6208 matched stmts if we cannot swap the non-matched ones.
6210 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6212 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6213 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6214 _mm_maskz_scalef_round_ss): New intrinsics.
6215 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6216 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6217 __builtin_ia32_scalefss_round): Remove.
6218 (__builtin_ia32_scalefsd_mask_round,
6219 __builtin_ia32_scalefss_mask_round): New intrinsics.
6220 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6221 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6222 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6223 "<round_constraint>")): Changed to ...
6224 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6225 "<round_scalar_constraint>")): ... this.
6226 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6227 %0, %1, %2<round_op3>}"): Changed to ...
6228 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6229 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6230 %2<round_scalar_mask_op3>}"): ... this.
6231 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6233 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6235 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6236 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6237 (_mm_maskz_sqrt_round_ss): New intrinsics.
6238 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6239 (__builtin_ia32_sqrtsd_mask_round)
6240 (__builtin_ia32_sqrtss_mask_round): New builtins.
6241 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6242 (__builtin_ia32_sqrtss_round): Remove.
6243 (__builtin_ia32_sqrtsd_mask_round)
6244 (__builtin_ia32_sqrtss_mask_round): New builtins.
6245 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6246 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6247 ((match_operand:VF_128 1 "vector_operand"
6248 "xBm,<round_constraint>")): Changed to ...
6249 ((match_operand:VF_128 1 "vector_operand"
6250 "xBm,<round_scalar_constraint>")): ... this.
6251 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6252 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6253 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6254 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6255 %<iptr>1<round_scalar_mask_op3>}): ... this.
6256 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6257 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6259 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6261 PR target/84266
6262 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6263 Cast vec_cmpeq result to correct type.
6264 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6265 Cast vec_cmpgt result to correct type.
6267 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6269 * final.c (final_scan_insn_1): Renamed from...
6270 (final_scan_insn): ... this. New wrapper, to recover
6271 seen from the outermost call in recursive ones.
6272 * config/sparc/sparc.c (output_return): Drop seen from call.
6273 (output_sibcall): Likewise.
6274 * config/visium/visium.c (output_branch): Likewise.
6276 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6278 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6279 function label.
6281 2018-02-10 Alan Modra <amodra@gmail.com>
6283 PR target/84300
6284 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6285 Specify LR as an input.
6287 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6289 PR sanitizer/83987
6290 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6291 remove_member_access_dummy_vars): New functions.
6292 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6293 lower_omp_1, execute_lower_omp): Use them.
6295 PR rtl-optimization/84308
6296 * shrink-wrap.c (spread_components): Release todo vector.
6298 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6300 PR rtl-optimization/57193
6301 * ira-color.c (struct allocno_color_data): Add member
6302 conflict_allocno_hard_prefs.
6303 (update_conflict_allocno_hard_prefs): New.
6304 (bucket_allocno_compare_func): Add a preference based on
6305 conflict_allocno_hard_prefs.
6306 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6307 (color_allocnos): Remove a dead code. Initiate
6308 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6310 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6312 PR target/84226
6313 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6314 constraint from =wa to wa. Avoid a subreg on the output operand,
6315 instead use a pseudo and subreg it in a move.
6316 (p9_xxbrd_<mode>): Changed to ...
6317 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6318 (p9_xxbrd_v2df): New expander.
6319 (p9_xxbrw_<mode>): Changed to ...
6320 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6321 (p9_xxbrw_v4sf): New expander.
6323 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6325 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6327 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6329 PR target/83926
6330 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6331 multiply in 32-bit mode.
6332 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6333 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6334 mode.
6336 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6338 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6339 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6340 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6341 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6343 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6345 PR lto/84213
6346 * dwarf2out.c (is_trivial_indirect_ref): New function.
6347 (dwarf2out_late_global_decl): Do not generate a location
6348 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6349 and that are not defined in the current unit.
6351 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6353 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6354 instead of a libcall for UNORDERED.
6356 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6358 PR target/82641
6359 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6360 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6362 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6364 PR target/PR84295
6365 * config/s390/s390.c (s390_set_current_function): Invoke
6366 s390_indirect_branch_settings also if fndecl didn't change.
6368 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6370 * config/rs6000/rs6000.md (blockage): Set length to zero.
6372 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6374 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6376 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6378 PR sanitizer/84285
6379 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6380 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6381 -static-lib*san.
6383 PR debug/84252
6384 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6385 PARALLEL incoming that failed vt_get_decl_and_offset check.
6387 PR middle-end/84237
6388 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6389 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6390 TREE_READONLY bit.
6391 (get_variable_section): For decls in named .bss* sections pass true as
6392 second argument to bss_initializer_p.
6394 2018-02-09 Marek Polacek <polacek@redhat.com>
6395 Jakub Jelinek <jakub@redhat.com>
6397 PR c++/83659
6398 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6399 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6400 Sync some changes from cxx_fold_indirect_ref.
6402 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6404 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6405 markers.
6406 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6407 (BLOCK_INLINE_ENTRY_LABEL): New.
6408 (dwarf2out_var_location): Disregard inline entry markers.
6409 (inline_entry_data): New struct.
6410 (inline_entry_data_hasher): New hashtable type.
6411 (inline_entry_data_hasher::hash): New.
6412 (inline_entry_data_hasher::equal): New.
6413 (inline_entry_data_table): New variable.
6414 (add_high_low_attributes): Add DW_AT_entry_pc and
6415 DW_AT_GNU_entry_view attributes if a pending entry is found
6416 in inline_entry_data_table. Add old entry_pc attribute only
6417 if debug nonbinding markers are disabled.
6418 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6419 markers are enabled.
6420 (block_within_block_p, dwarf2out_inline_entry): New.
6421 (dwarf2out_finish): Check that no entries remained in
6422 inline_entry_data_table.
6423 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6424 (final_scan_insn, notice_source_line): Likewise.
6425 (rest_of_clean_state): Skip inline entry markers.
6426 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6427 markers.
6428 * gimple.c (gimple_build_debug_inline_entry): New.
6429 * gimple.h (enum gimple_debug_subcode): Add
6430 GIMPLE_DEBUG_INLINE_ENTRY.
6431 (gimple_build_debug_inline_entry): Declare.
6432 (gimple_debug_inline_entry_p): New.
6433 (gimple_debug_nonbind_marker_p): Adjust.
6434 * insn-notes.def (INLINE_ENTRY): New.
6435 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6436 inline entry marker notes.
6437 (print_insn): Likewise.
6438 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6439 (INSN_DEBUG_MARKER_KIND): Likewise.
6440 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6441 * tree-inline.c (expand_call_inline): Build and insert
6442 debug_inline_entry stmt.
6443 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6444 inline entry blocks early, if nonbind markers are enabled.
6445 (dump_scope_block): Dump fragment info.
6446 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6447 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6448 (gimple_build_debug_inline_entry): New.
6449 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6450 Enable/disable inline entry points too.
6451 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6452 (DEBUG_INSN): Describe inline entry markers.
6454 * common.opt (gvariable-location-views): New.
6455 (gvariable-location-views=incompat5): New.
6456 * config.in: Rebuilt.
6457 * configure: Rebuilt.
6458 * configure.ac: Test assembler for view support.
6459 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6460 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6461 * dwarf2out.c (var_loc_view): New typedef.
6462 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6463 (dwarf2out_locviews_in_attribute): New.
6464 (dwarf2out_locviews_in_loclist): New.
6465 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6466 (enum dw_line_info_opcode): Add LI_adv_address.
6467 (struct dw_line_info_table): Add view.
6468 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6469 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6470 (zero_view_p): New variable.
6471 (ZERO_VIEW_P): New macro.
6472 (output_asm_line_debug_info): New.
6473 (struct var_loc_node): Add view.
6474 (add_AT_view_list, AT_loc_list): New.
6475 (add_var_loc_to_decl): Add view param. Test it against last.
6476 (new_loc_list): Add view params. Record them.
6477 (AT_loc_list_ptr): Handle loc and view lists.
6478 (view_list_to_loc_list_val_node): New.
6479 (print_dw_val): Handle dw_val_class_view_list.
6480 (size_of_die): Likewise.
6481 (value_format): Likewise.
6482 (loc_list_has_views): New.
6483 (gen_llsym): Set vl_symbol too.
6484 (maybe_gen_llsym, skip_loc_list_entry): New.
6485 (dwarf2out_maybe_output_loclist_view_pair): New.
6486 (output_loc_list): Output view list or entries too.
6487 (output_view_list_offset): New.
6488 (output_die): Handle dw_val_class_view_list.
6489 (output_dwarf_version): New.
6490 (output_compilation_unit_header): Use it.
6491 (output_skeleton_debug_sections): Likewise.
6492 (output_rnglists, output_line_info): Likewise.
6493 (output_pubnames, output_aranges): Update version comments.
6494 (output_one_line_info_table): Output view numbers in asm comments.
6495 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6496 Call maybe_gen_llsym.
6497 (loc_list_from_tree_1): Adjust.
6498 (add_AT_location_description): Create view list attribute if
6499 needed, check it's absent otherwise.
6500 (convert_cfa_to_fb_loc_list): Adjust.
6501 (maybe_emit_file): Call output_asm_line_debug_info for test.
6502 (dwarf2out_var_location): Reset views as needed. Precompute
6503 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6504 attribute. Set view.
6505 (new_line_info_table): Reset next view.
6506 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6507 (dwarf2out_source_line): Likewise. Output view resets and labels to
6508 the assembler, or select appropriate line info opcodes.
6509 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6510 (optimize_string_length): Catch it. Adjust.
6511 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6512 dw_val_class_view_list, and remove it if no longer needed.
6513 (hash_loc_list): Hash view numbers.
6514 (loc_list_hasher::equal): Compare them.
6515 (optimize_location_lists): Check whether a view list symbol is
6516 needed, and whether the locview attribute is present, and
6517 whether they match. Remove the locview attribute if no longer
6518 needed.
6519 (index_location_lists): Call skip_loc_list_entry for test.
6520 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6521 Use output_dwarf_version.
6522 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6523 (struct dw_val_node): Add val_view_list.
6524 * final.c (SEEN_NEXT_VIEW): New.
6525 (set_next_view_needed): New.
6526 (clear_next_view_needed): New.
6527 (maybe_output_next_view): New.
6528 (final_start_function): Rename to...
6529 (final_start_function_1): ... this. Take pointer to FIRST,
6530 add SEEN parameter. Emit param bindings in the initial view.
6531 (final_start_function): Reintroduce SEEN-less interface.
6532 (final): Rename to...
6533 (final_1): ... this. Take SEEN parameter. Output final pending
6534 next view at the end.
6535 (final): Reintroduce seen-less interface.
6536 (final_scan_insn): Output pending next view before switching
6537 sections or ending a block. Mark the next view as needed when
6538 outputting variable locations. Notify debug backend of section
6539 changes, and of location view changes.
6540 (rest_of_handle_final): Adjust.
6541 * toplev.c (process_options): Autodetect value for debug variable
6542 location views option. Warn on incompat5 without -gdwarf-5.
6543 * doc/invoke.texi (gvariable-location-views): New.
6544 (gvariable-location-views=incompat5): New.
6545 (gno-variable-location-views): New.
6547 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6549 PR tree-optimization/84136
6550 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6551 that the result of find_edge is non-NULL.
6553 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6555 PR target/83008
6556 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6557 storing integer register in SImode. Fix cost of 256 and 512
6558 byte aligned SSE register store.
6560 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6562 * config/i386/i386.c (ix86_multiplication_cost): Fix
6563 multiplication cost for TARGET_AVX512DQ.
6565 2018-02-08 Marek Polacek <polacek@redhat.com>
6567 PR tree-optimization/84238
6568 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6569 get_range_strlen.
6571 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6573 PR tree-optimization/84265
6574 * tree-vect-stmts.c (vectorizable_store): Don't treat
6575 VMAT_CONTIGUOUS accesses as grouped.
6576 (vectorizable_load): Likewise.
6578 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6580 PR tree-optimization/81635
6581 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6582 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6583 (test_round_for_mask): New functions.
6584 (wide_int_cc_tests): Call test_round_for_mask.
6585 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6586 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6587 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6588 range returned by get_range_info.
6590 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6592 PR ipa/81360
6593 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6594 * symtab.c: Include builtins.h
6595 (symtab_node::output_to_lto_symbol_table_p): Move here
6596 from lto-streamer-out.c:output_symbol_p.
6597 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6598 (output_symbol_p): Move all logic to symtab.c
6599 (produce_symtab): Update.
6601 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6603 * config/s390/s390-opts.h (enum indirect_branch): Define.
6604 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6605 (s390_indirect_branch_via_thunk)
6606 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6607 (enum s390_indirect_branch_type): Define.
6608 * config/s390/s390.c (struct s390_frame_layout, struct
6609 machine_function): Remove.
6610 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6611 (indirect_branch_table_label_no, indirect_branch_table_name):
6612 Define variables.
6613 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6614 (enum s390_indirect_branch_option): Define.
6615 (s390_return_addr_from_memory): New function.
6616 (s390_handle_string_attribute): New function.
6617 (s390_attribute_table): Add new attribute handler.
6618 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6619 (s390_indirect_branch_via_thunk): New function.
6620 (s390_indirect_branch_via_inline_thunk): New function.
6621 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6622 sibling call optimization for non z10 compiles.
6623 (s390_emit_call): Force indirect branch target to be a single
6624 register. Add r1 clobber for non-z10 compiles.
6625 (s390_emit_epilogue): Emit return jump via return_use expander.
6626 (s390_reorg): Handle JUMP_INSNs as execute targets.
6627 (s390_option_override_internal): Perform validity checks for the
6628 new command line options.
6629 (s390_indirect_branch_attrvalue): New function.
6630 (s390_indirect_branch_settings): New function.
6631 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6632 (s390_output_indirect_thunk_function): New function.
6633 (s390_code_end): Implement target hook.
6634 (s390_case_values_threshold): Implement target hook.
6635 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6636 macros.
6637 * config/s390/s390.h (struct s390_frame_layout)
6638 (struct machine_function): Move here from s390.c.
6639 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6640 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6641 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6642 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6643 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6644 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6645 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6646 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6647 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6648 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6649 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6650 (mnemonic attribute): Add values which aren't recognized
6651 automatically.
6652 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6653 pattern for branch conversion. Fix mnemonic attribute.
6654 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6655 indirect branch via thunk if requested.
6656 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6657 ("*indirect_jump"): Disable for branch conversion using out of
6658 line thunks.
6659 ("indirect_jump_via_thunk<mode>_z10")
6660 ("indirect_jump_via_thunk<mode>")
6661 ("indirect_jump_via_inlinethunk<mode>_z10")
6662 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6663 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6664 ("casesi_jump_via_inlinethunk<mode>_z10")
6665 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6666 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6667 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6668 ("*indirect2_jump"): Disable for branch conversion.
6669 ("casesi_jump"): Turn into expander and expand patterns for branch
6670 conversion.
6671 ("return_use"): New expander.
6672 ("*return"): Emit return via thunk and rename it to ...
6673 ("*return<mode>"): ... this one.
6674 * config/s390/s390.opt: Add new options and and enum for the
6675 option values.
6677 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6679 * lra-constraints.c (match_reload): Unconditionally use
6680 gen_lowpart_SUBREG, rather than selecting between that
6681 and equivalent gen_rtx_SUBREG code.
6683 2018-02-08 Richard Biener <rguenther@suse.de>
6685 PR tree-optimization/84233
6686 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6687 changed flag instead of boguously re-using phi_inserted.
6689 2018-02-08 Martin Jambor <mjambor@suse.cz>
6691 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6692 static local variables.
6694 2018-02-08 Richard Biener <rguenther@suse.de>
6696 PR tree-optimization/84278
6697 * tree-vect-stmts.c (vectorizable_store): When looking for
6698 smaller vector types to perform grouped strided loads/stores
6699 make sure the mode is supported by the target.
6700 (vectorizable_load): Likewise.
6702 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6704 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6705 Increase LDP/STP opportunities by adding adjacent callee-saves.
6707 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6709 PR rtl-optimization/84068
6710 PR rtl-optimization/83459
6711 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6713 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6715 PR tree-optimization/84224
6716 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6717 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6718 non-zero arguments.
6720 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6722 PR target/84113
6723 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6724 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6725 count, remove one USE.
6727 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6729 * doc/install.texi (Configuration): Document the
6730 --with-long-double-format={ibm,ieee} PowerPC configuration
6731 options.
6733 PR target/84154
6734 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6735 Convert from define_expand to be define_insn_and_split. Rework
6736 float/double/_Float128 conversions to QI/HI/SImode to work with
6737 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6738 conversions to QI/HImode types did a store and then a load to
6739 truncate the value. For conversions to VSX registers, don't split
6740 the insn, instead emit the code directly. Use the code iterator
6741 any_fix to combine signed and unsigned conversions.
6742 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6743 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6744 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6745 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6746 (fix_<mode>di2_hw): Likewise.
6747 (fixuns_<mode>di2_hw): Likewise.
6748 (fix_<mode>si2_hw): Likewise.
6749 (fixuns_<mode>si2_hw): Likewise.
6750 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6751 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6752 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6753 fix<uns>_trunc<SFDF:mode>si2_p8.
6754 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6755 used.
6756 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6757 (fix<uns>_<mode>_mem): Likewise.
6758 (fctiw<u>z_<mode>_mem): Likewise.
6759 (fix<uns>_<mode>_mem): Likewise.
6760 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6761 the register allocator from doing a direct move to the GPRs to do
6762 a store, and instead use the ISA 3.0 store byte/half-word from
6763 vector register instruction. For IEEE 128-bit floating point,
6764 also optimize stores of 32-bit ints.
6765 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6767 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6769 * genextract.c (push_pathstr_operand): New function to support
6770 [a-zA-Z].
6771 (walk_rtx): Call push_pathstr_operand.
6772 (print_path): Support [a-zA-Z].
6774 2018-02-07 Richard Biener <rguenther@suse.de>
6776 PR tree-optimization/84037
6777 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6778 (cse_and_gimplify_to_preheader): Declare.
6779 (vect_get_place_in_interleaving_chain): Likewise.
6780 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6781 ivexpr_map.
6782 (_loop_vec_info::~_loop_vec_info): Delete it.
6783 (cse_and_gimplify_to_preheader): New function.
6784 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6785 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6786 (vectorizable_load): Likewise. For grouped stores always base
6787 the IV on the first element.
6788 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6789 condition before gimplifying.
6791 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6793 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6794 *DIV_EXPR and *MOD_EXPR.
6796 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6798 PR target/84248
6799 * config/i386/i386.c (ix86_option_override_internal): Mask out
6800 the CF_SET bit when checking -fcf-protection.
6802 2018-02-07 Tom de Vries <tom@codesourcery.com>
6804 PR libgomp/84217
6805 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6806 enough.
6808 2018-02-07 Richard Biener <rguenther@suse.de>
6810 PR tree-optimization/84204
6811 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6812 this place.
6814 PR tree-optimization/84205
6815 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6816 special-case isl_ast_op_zdiv_r.
6818 PR tree-optimization/84223
6819 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6820 Only add conditions from within the region.
6821 (gather_bbs::after_dom_children): Adjust.
6823 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6825 PR target/84209
6826 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6827 * config/avr/avr.md: Only post-reload split REG-REG moves if
6828 either register is GENERAL_REG_P.
6830 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6832 PR tree-optimization/84235
6833 * tree-ssa-scopedtables.c
6834 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6835 if the subtraction is performed in floating point type where NaNs are
6836 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6837 build 1. Formatting fix.
6839 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6841 PR target/84146
6842 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6843 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6844 and skip it regardless of bb boundaries. Use CALL_P macro,
6845 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6846 unnecessarily, formatting fix.
6848 2018-02-06 Michael Collison <michael.collison@arm.com>
6850 * config/arm/thumb2.md:
6851 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6852 (*thumb_mov_notscc): Ditto.
6854 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6856 PR target/84154
6857 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6858 unsigned_fix, not "s".
6860 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6862 * configure.ac (gcc_fn_eh_frame_ro): New function.
6863 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6864 correct .eh_frame permissions.
6865 * configure: Regenerate.
6867 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6869 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6870 irrelevant options.
6872 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6874 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6875 Display warning message for -mno-speculate-indirect-jumps.
6877 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6879 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6880 Undocumented.
6881 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6883 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6885 PR tree-optimization/84225
6886 * tree-eh.c (find_trapping_overflow): Only call
6887 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6889 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6891 PR target/84145
6892 * config/i386/i386.c: Reimplement the check of possible options
6893 -mibt/-mshstk conbination. Change error messages.
6894 * doc/invoke.texi: Fix a typo: remove extra '='.
6896 2018-02-06 Marek Polacek <polacek@redhat.com>
6898 PR tree-optimization/84228
6899 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6901 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6903 PR target/82641
6904 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6905 emitted arch directives.
6906 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6907 __ARM_FEATURE_COPROC before changing architectures.
6909 2018-02-06 Richard Biener <rguenther@suse.de>
6911 * config/i386/i386.c (print_reg): Fix typo.
6912 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6914 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6916 * configure: Regenerate.
6918 2018-02-05 Martin Sebor <msebor@redhat.com>
6920 PR tree-optimization/83369
6921 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6922 inlining context.
6924 2018-02-05 Martin Liska <mliska@suse.cz>
6926 * doc/invoke.texi: Cherry-pick upstream r323995.
6928 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6930 * ira.c (ira_init_register_move_cost): Adjust comment.
6932 2018-02-05 Martin Liska <mliska@suse.cz>
6934 PR gcov-profile/84137
6935 * doc/gcov.texi: Fix typo in documentation.
6937 2018-02-05 Martin Liska <mliska@suse.cz>
6939 PR gcov-profile/83879
6940 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6941 using dlopen functionality.
6943 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6945 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6946 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6947 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6948 _mm_maskz_range_round_ss): New intrinsics.
6949 (__builtin_ia32_rangesd128_round)
6950 (__builtin_ia32_rangess128_round): Remove.
6951 (__builtin_ia32_rangesd128_mask_round,
6952 __builtin_ia32_rangess128_mask_round): New builtins.
6953 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6954 __builtin_ia32_rangess128_round): Remove.
6955 (__builtin_ia32_rangesd128_mask_round,
6956 __builtin_ia32_rangess128_mask_round): New builtins.
6957 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6958 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6959 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6960 "<round_saeonly_constraint>")): Changed to ...
6961 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6962 "<round_saeonly_scalar_constraint>")): ... this.
6963 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6964 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6965 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6966 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6967 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6969 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6971 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6972 options.
6973 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6974 Remove all values except native, 8540 and 8548.
6976 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6978 * config/i386/i386.c (ix86_output_function_return): Pass
6979 INVALID_REGNUM, instead of -1, as invalid register number to
6980 indirect_thunk_name and output_indirect_thunk.
6982 2018-02-02 Julia Koval <julia.koval@intel.com>
6984 * config.gcc: Add -march=icelake.
6985 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6986 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6987 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6988 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6989 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6990 (processor_target_table): Add icelake.
6991 (ix86_option_override_internal): Handle new PTAs.
6992 (get_builtin_code_for_version): Handle icelake.
6993 (M_INTEL_COREI7_ICELAKE): New.
6994 (fold_builtin_cpu): Handle icelake.
6995 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6996 * doc/invoke.texi: Add -march=icelake.
6998 2018-02-02 Julia Koval <julia.koval@intel.com>
7000 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7001 to wide_int_bitmask.
7002 * wide-int-bitmask.h: New.
7004 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7006 PR target/84066
7007 * config/i386/i386.md: Replace Pmode with word_mode in
7008 builtin_setjmp_setup and builtin_longjmp to support x32.
7010 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7012 PR target/56010
7013 PR target/83743
7014 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7015 #include "opts.h".
7016 (rs6000_supported_cpu_names): New static variable.
7017 (linux_cpu_translation_table): Likewise.
7018 (elf_platform) <cpu>: Define new static variable and use it.
7019 Translate kernel AT_PLATFORM name to canonical name if needed.
7020 Error if platform name is unknown.
7022 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7024 PR target/84089
7025 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7027 2018-02-01 Jeff Law <law@redhat.com>
7029 PR target/84128
7030 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7031 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7032 the scratch if RELEASE_VIA_POP is false.
7033 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7034 If we have to save a temporary register, decrement SIZE appropriately.
7035 Pass new arguments to release_scratch_register_on_entry.
7036 (ix86_adjust_stack_and_probe): Likewise.
7037 (ix86_emit_probe_stack_range): Pass new arguments to
7038 release_scratch_register_on_entry.
7040 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7042 PR rtl-optimization/84157
7043 * combine.c (change_zero_ext): Use REG_P predicate in
7044 front of HARD_REGISTER_P predicate.
7046 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7048 * config/avr/avr.c (avr_option_override): Move disabling of
7049 -fdelete-null-pointer-checks to...
7050 * common/config/avr/avr-common.c (avr_option_optimization_table):
7051 ...here.
7053 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7055 PR tree-optimization/81635
7056 * tree-data-ref.c (split_constant_offset_1): For types that
7057 wrap on overflow, try to use range info to prove that wrapping
7058 cannot occur.
7060 2018-02-01 Renlin Li <renlin.li@arm.com>
7062 PR target/83370
7063 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7064 TAILCALL_ADDR_REGS.
7065 (aarch64_register_move_cost): Likewise.
7066 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7067 TAILCALL_ADDR_REGS.
7068 (REG_CLASS_NAMES): Likewise.
7069 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7070 TAILCALL_ADDR_REGS. Remove IP registers.
7071 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7073 2018-02-01 Richard Biener <rguenther@suse.de>
7075 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7076 for specifying RPO order and allow NULL for that.
7077 * domwalk.c (dom_walker::dom_walker): Likewise.
7078 (dom_walker::walk): Handle NULL RPO order.
7079 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7080 in RPO order.
7081 (rewrite_update_dom_walker): Likewise.
7082 (mark_def_dom_walker): Likewise.
7084 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7086 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7087 (aarch64_maybe_expand_sve_subreg_move): Declare.
7088 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7089 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7090 predicate.
7091 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7092 that are semantically a reverse operation.
7093 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7094 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7095 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7096 functions.
7097 (aarch64_can_change_mode_class): For big-endian, forbid changes
7098 between two SVE modes if they have different element sizes.
7100 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7102 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7103 the TImode handling for big-endian targets.
7105 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7107 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7108 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7109 not just bytes.
7110 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7111 Remove BSWAP handing for big-endian targets and use the form of
7112 LD1RQ appropariate for the mode.
7114 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7116 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7117 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7118 duplicated element.
7120 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7122 PR tearget/83845
7123 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7124 check for operands that need to go through aarch64_sve_reload_be.
7126 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7128 PR tree-optimization/81661
7129 PR tree-optimization/84117
7130 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7131 * tree-eh.c: Include gimplify.h.
7132 (find_trapping_overflow, replace_trapping_overflow,
7133 rewrite_to_non_trapping_overflow): New functions.
7134 * tree-vect-loop.c: Include tree-eh.h.
7135 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7136 * tree-data-ref.c: Include tree-eh.h.
7137 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7139 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7141 PR rtl-optimization/84123
7142 * combine.c (change_zero_ext): Check if hard register satisfies
7143 can_change_dest_mode before calling gen_lowpart_SUBREG.
7145 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7147 PR target/82444
7148 * ira.c (ira_init_register_move_cost): Remove assert.
7150 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7152 PR rtl-optimization/84071
7153 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7154 * doc/tm.texi: Regenerate.
7156 2018-01-31 Richard Biener <rguenther@suse.de>
7158 PR tree-optimization/84132
7159 * tree-data-ref.c (analyze_miv_subscript): Properly
7160 check whether evolution_function_is_affine_multivariate_p
7161 before calling gcd_of_steps_may_divide_p.
7163 2018-01-31 Julia Koval <julia.koval@intel.com>
7165 PR target/83618
7166 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7167 * config/i386/i386.md (rdpid_rex64) New.
7168 (rdpid): Make 32bit only.
7170 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7172 PR lto/84105
7173 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7174 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7176 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7178 Revert
7179 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7181 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7183 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7185 PR rtl-optimization/84071
7186 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7187 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7189 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7191 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7192 (arc_attribute_table): Add 'aux' attribute.
7193 (arc_in_small_data_p): Consider aux like variables.
7194 (arc_is_aux_reg_p): New function.
7195 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7196 (arc_get_aux_arg): New function.
7197 (prepare_move_operands): Handle aux-register access.
7198 (arc_handle_aux_attribute): New function.
7199 * doc/extend.texi (ARC Variable attributes): Add subsection.
7201 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7203 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7204 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7205 (arc_attribute_table): Add 'uncached' attribute.
7206 (arc_print_operand): Print '.di' flag for uncached memory
7207 accesses.
7208 (arc_in_small_data_p): Do not consider for small data the uncached
7209 types.
7210 (arc_is_uncached_mem_p): New function.
7211 * config/arc/predicates.md (compact_store_memory_operand): Check
7212 for uncached memory accesses.
7213 (nonvol_nonimm_operand): Likewise.
7214 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7216 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7218 PR c/84100
7219 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7220 falign-loops=): Add Optimization flag.
7222 2018-01-30 Jeff Law <law@redhat.com>
7224 PR target/84064
7225 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7226 INT_REGISTERS_SAVED. Check it prior to calling
7227 get_scratch_register_on_entry.
7228 (ix86_adjust_stack_and_probe): Similarly.
7229 (ix86_emit_probe_stack_range): Similarly.
7230 (ix86_expand_prologue): Corresponding changes.
7232 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7234 PR target/40411
7235 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7236 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7238 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7240 PR target/84112
7241 * lra-constraints.c (curr_insn_transform): Process AND in the
7242 address.
7244 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7246 PR rtl-optimization/83986
7247 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7248 dependence against last_pending_memory_flush in addition to
7249 pending_jump_insns.
7251 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7253 PR tree-optimization/81611
7254 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7255 copies.
7257 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7259 PR target/83758
7260 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7261 a reg rtx.
7263 2018-01-30 Richard Biener <rguenther@suse.de>
7264 Jakub Jelinek <jakub@redhat.com>
7266 PR tree-optimization/84111
7267 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7268 inner loops added during recursion, as they don't have up-to-date
7269 SSA form.
7271 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7273 PR ipa/81360
7274 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7275 (can_inline_edge_by_limits_p): ... here.
7276 (can_early_inline_edge_p, check_callers,
7277 update_caller_keys, update_callee_keys, recursive_inlining,
7278 add_new_edges_to_heap, speculation_useful_p,
7279 inline_small_functions,
7280 inline_small_functions, flatten_function,
7281 inline_to_all_callers_1): Update.
7283 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7285 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7286 zeros correctly.
7288 2018-01-30 Richard Biener <rguenther@suse.de>
7290 PR tree-optimization/83008
7291 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7292 invariant and constant vector uses in stmts when they need
7293 more than one stmt.
7295 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7297 PR bootstrap/84017
7298 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7299 * configure: Regenerate.
7301 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7303 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7304 pattern.
7305 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7306 Use gen_rtx_REG rather than gen_lowpart.
7308 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7310 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7311 rather than 0 when creating partial subregs.
7313 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7315 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7316 of usage.
7318 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7320 PR target/81550
7321 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7322 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7323 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7324 flags. This restores the settings used before the 2017-07-24.
7325 Turning off pre increment/decrement/modify allows IVOPTS to
7326 optimize DF/SF loops where the index is an int.
7328 2018-01-29 Richard Biener <rguenther@suse.de>
7329 Kelvin Nilsen <kelvin@gcc.gnu.org>
7331 PR bootstrap/80867
7332 * tree-vect-stmts.c (vectorizable_call): Don't call
7333 targetm.vectorize_builtin_md_vectorized_function if callee is
7334 NULL.
7336 2018-01-22 Carl Love <cel@us.ibm.com>
7338 * doc/extend.tex: Fix typo in second arg in
7339 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7341 2018-01-29 Richard Biener <rguenther@suse.de>
7343 PR tree-optimization/84086
7344 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7345 (flush_ssaname_freelist): When SSA names were released reset
7346 the SCEV hash table.
7348 2018-01-29 Richard Biener <rguenther@suse.de>
7350 PR tree-optimization/84057
7351 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7352 removed paths when removing edges.
7354 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7356 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7357 -mfunction-return=@var{choice}.
7359 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7361 PR diagnostic/84034
7362 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7363 Handle CR like TAB.
7364 (layout::print_source_line): Likewise.
7365 (test_get_line_width_without_trailing_whitespace): Add test cases.
7367 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7369 PR middle-end/84040
7370 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7371 debug insns.
7373 2018-01-26 Jim Wilson <jimw@sifive.com>
7375 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7377 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7378 specified.
7380 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7382 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7383 and CMP + SUB-immediate -> SUBS.
7385 2018-01-26 Martin Sebor <msebor@redhat.com>
7387 PR tree-optimization/83896
7388 * tree-ssa-strlen.c (get_string_len): Rename...
7389 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7390 Avoid assuming length is constant.
7391 (handle_char_store): Use HOST_WIDE_INT for string length.
7393 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7395 PR target/81763
7396 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7397 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7399 2018-01-26 Richard Biener <rguenther@suse.de>
7401 PR rtl-optimization/84003
7402 * dse.c (record_store): Only record redundant stores when
7403 the earlier store aliases at least all accesses the later one does.
7405 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7407 PR rtl-optimization/83985
7408 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7409 REG_CFA_RESTORE insns.
7410 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7411 REG_CFA_RESTORE insns here.
7413 PR c/83989
7414 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7415 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7417 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7419 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7420 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7421 (arc_init): Likewise.
7422 (arc_override_options): Likewise.
7423 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7424 value.
7425 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7426 support.
7427 * config/arc/arc.h (TARGET_DBNZ): Define.
7428 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7429 properly set the tune attribute.
7430 (dbnz): Use TARGET_DBNZ guard.
7431 * config/arc/arc.opt (mtune): Add core3 option.
7433 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7435 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7436 recognize new pic like addresses.
7437 (arc_delegitimize_address): Clean up.
7439 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7441 * config/arc/arc-arches.def: Option mrf16 valid for all
7442 architectures.
7443 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7444 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7445 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7446 * config/arc/arc-tables.opt: Regenerate.
7447 * config/arc/arc.c (arc_conditional_register_usage): Handle
7448 reduced register file case.
7449 (arc_file_start): Set must have build attributes.
7450 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7451 mrf16 option value.
7452 * config/arc/arc.opt (mrf16): Add new option.
7453 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7454 * config/arc/genmultilib.awk: Handle new mrf16 option.
7455 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7456 * config/arc/t-multilib: Regenerate.
7457 * doc/invoke.texi (ARC Options): Document mrf16 option.
7459 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7461 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7462 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7463 (arc_attribute_table): Add 'secure_call' attribute.
7464 (arc_print_operand): Print secure call operand.
7465 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7466 secure.
7467 (arc_is_secure_call_p): New function. * config/arc/arc.md
7468 (call_i): Add support for sjli instruction.
7469 (call_value_i): Likewise.
7470 * config/arc/constraints.md (Csc): New constraint.
7472 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7473 John Eric Martin <John.Martin@emmicro-us.com>
7475 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7476 * config/arc/arc.c (_arc_jli_section): New struct.
7477 (arc_jli_section): New type.
7478 (rc_jli_sections): New static variable.
7479 (arc_handle_jli_attribute): New function.
7480 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7481 (arc_file_end): New function.
7482 (TARGET_ASM_FILE_END): Define.
7483 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7484 (arc_add_jli_section): New function.
7485 (jli_call_scan): Likewise.
7486 (arc_reorg): Call jli_call_scan.
7487 (arc_output_addsi): Remove 'S' from printing asm operand.
7488 (arc_is_jli_call_p): New function.
7489 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7490 operand.
7491 (movhi_insn): Likewise.
7492 (movsi_insn): Likewise.
7493 (movsi_set_cc_insn): Likewise.
7494 (loadqi_update): Likewise.
7495 (load_zeroextendqisi_update): Likewise.
7496 (load_signextendqisi_update): Likewise.
7497 (loadhi_update): Likewise.
7498 (load_zeroextendhisi_update): Likewise.
7499 (load_signextendhisi_update): Likewise.
7500 (loadsi_update): Likewise.
7501 (loadsf_update): Likewise.
7502 (movsicc_insn): Likewise.
7503 (bset_insn): Likewise.
7504 (bxor_insn): Likewise.
7505 (bclr_insn): Likewise.
7506 (bmsk_insn): Likewise.
7507 (bicsi3_insn): Likewise.
7508 (cmpsi_cc_c_insn): Likewise.
7509 (movsi_ne): Likewise.
7510 (movsi_cond_exec): Likewise.
7511 (clrsbsi2): Likewise.
7512 (norm_f): Likewise.
7513 (normw): Likewise.
7514 (swap): Likewise.
7515 (divaw): Likewise.
7516 (flag): Likewise.
7517 (sr): Likewise.
7518 (kflag): Likewise.
7519 (ffs): Likewise.
7520 (ffs_f): Likewise.
7521 (fls): Likewise.
7522 (call_i): Remove 'S' asm letter, add jli instruction.
7523 (call_value_i): Likewise.
7524 * config/arc/arc.op (mjli-always): New option.
7525 * config/arc/constraints.md (Cji): New constraint.
7526 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7527 operand.
7528 (subsf3_fpx): Likewise.
7529 (mulsf3_fpx): Likewise.
7530 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7531 asm operand.
7532 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7533 function attrbutes.
7534 * doc/invoke.texi (ARC): Document mjli-always option.
7536 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7538 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7539 avoid addition with 0 and use incw and decw where possible.
7541 2018-01-26 Richard Biener <rguenther@suse.de>
7543 PR tree-optimization/81082
7544 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7545 association if it requires casting to unsigned.
7546 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7547 from fold_plusminus_mult_expr to catch important cases late when
7548 range info is available.
7550 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7552 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7553 * configure.ac (hidden_linkonce): New test.
7554 * configure: Regenerate.
7555 * config.in: Regenerate.
7557 2018-01-26 Julia Koval <julia.koval@intel.com>
7559 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7560 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7561 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7562 _mm_mask_bitshuffle_epi64_mask): Fix type.
7563 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7564 USI_FTYPE_V4DI_V4DI_USI): Remove.
7565 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7566 __builtin_ia32_vpshufbitqmb256_mask,
7567 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7568 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7569 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7571 2018-01-26 Alan Modra <amodra@gmail.com>
7573 PR target/84033
7574 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7575 UNSPEC_VBPERMQ. Sort other unspecs.
7577 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7579 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7581 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7583 PR middle-end/83055
7584 * predict.c (drop_profile): Do not push/pop cfun; update also
7585 node->count.
7586 (handle_missing_profiles): Fix logic looking for zero profiles.
7588 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7590 PR middle-end/83977
7591 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7592 on functions with #pragma omp declare simd or functions with simd
7593 attribute.
7594 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7595 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7596 Remove trailing \n from warning_at calls.
7598 2018-01-25 Tom de Vries <tom@codesourcery.com>
7600 PR target/84028
7601 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7602 for neutered workers.
7604 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7606 PR target/68467
7607 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7608 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7610 2018-01-24 Jeff Law <law@redhat.com>
7612 PR target/83994
7613 * i386.c (get_probe_interval): Move to earlier point.
7614 (ix86_compute_frame_layout): If -fstack-clash-protection and
7615 the frame is larger than the probe interval, then use pushes
7616 to save registers rather than reg->mem moves.
7617 (ix86_expand_prologue): Remove conditional for int_registers_saved
7618 assertion.
7620 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7622 PR target/84014
7623 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7624 min/max for never referenced object.
7626 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7628 PR middle-end/83977
7629 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7630 here.
7631 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7632 attributes from DECL_ATTRIBUTES (decl) without affecting
7633 DECL_ATTRIBUTES (current_function_decl).
7634 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7635 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7637 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7639 PR tree-optimization/83979
7640 * fold-const.c (fold_comparison): Use constant_boolean_node
7641 instead of boolean_{true,false}_node.
7643 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7645 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7646 with zero counts.
7648 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7650 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7651 Simplify the clause that sets the length attribute.
7652 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7653 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7654 clause that sets the length attribute.
7655 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7657 2018-01-24 Tom de Vries <tom@codesourcery.com>
7659 PR target/83589
7660 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7661 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7662 Add strict parameter.
7663 (prevent_branch_around_nothing): Insert dummy insn between branch to
7664 label and label with no ptx insn inbetween.
7665 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7667 2018-01-24 Tom de Vries <tom@codesourcery.com>
7669 PR target/81352
7670 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7671 for neutered threads in warp.
7672 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7674 2018-01-24 Richard Biener <rguenther@suse.de>
7676 PR tree-optimization/83176
7677 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7678 operands.
7680 2018-01-24 Richard Biener <rguenther@suse.de>
7682 PR tree-optimization/82819
7683 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7684 code generating pluses that are no-ops in the target precision.
7686 2018-01-24 Richard Biener <rguenther@suse.de>
7688 PR middle-end/84000
7689 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7691 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7693 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7694 to merge probabilities.
7695 * predict.c (probably_never_executed): Also mark as cold functions
7696 with global 0 profile and guessed local profile.
7697 * profile-count.c (profile_probability::combine_with_count): New
7698 member function.
7699 * profile-count.h (profile_probability::operator*,
7700 profile_probability::operator*=, profile_probability::operator/,
7701 profile_probability::operator/=): Reduce precision to adjusted
7702 and set value to guessed on contradictory divisions.
7703 (profile_probability::combine_with_freq): Remove.
7704 (profile_probability::combine_wiht_count): Declare.
7705 (profile_count::force_nonzero):: Set to adjusted.
7706 (profile_count::probability_in):: Set quality to adjusted.
7707 * tree-ssa-tail-merge.c (replace_block_by): Use
7708 combine_with_count.
7710 2018-01-23 Andrew Waterman <andrew@sifive.com>
7711 Jim Wilson <jimw@sifive.com>
7713 * config/riscv/riscv.c (riscv_stack_boundary): New.
7714 (riscv_option_override): Set riscv_stack_boundary. Handle
7715 riscv_preferred_stack_boundary_arg.
7716 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7717 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7718 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7719 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7720 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7721 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7723 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7725 PR target/83905
7726 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7727 of struct ix86_frame.
7728 (ix86_expand_epilogue): Likewise. Add a local variable for
7729 the reg_save_offset field in struct ix86_frame.
7731 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7733 PR tree-optimization/82604
7734 * tree-loop-distribution.c (enum partition_kind): New enum item
7735 PKIND_PARTIAL_MEMSET.
7736 (partition_builtin_p): Support above new enum item.
7737 (generate_code_for_partition): Ditto.
7738 (compute_access_range): Differentiate cases that equality can be
7739 proven at all loops, the innermost loops or no loops.
7740 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7741 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7742 (finalize_partitions, distribute_loop): Don't fuse partition of
7743 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7744 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7745 parloop is enabled.
7747 2018-01-23 Martin Liska <mliska@suse.cz>
7749 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7750 order to ignore the predictor.
7751 (PRED_POLYMORPHIC_CALL): Likewise.
7752 (PRED_RECURSIVE_CALL): Likewise.
7754 2018-01-23 Martin Liska <mliska@suse.cz>
7756 * tree-profile.c (tree_profiling): Print function header to
7757 aware reader which function we are working on.
7758 * value-prof.c (gimple_find_values_to_profile): Do not print
7759 not interesting value histograms.
7761 2018-01-23 Martin Liska <mliska@suse.cz>
7763 * profile-count.h (enum profile_quality): Add
7764 profile_uninitialized as the first value. Do not number values
7765 as they are zero based.
7766 (profile_count::verify): Update sanity check.
7767 (profile_probability::verify): Likewise.
7769 2018-01-23 Nathan Sidwell <nathan@acm.org>
7771 * doc/invoke.texi (ffor-scope): Deprecate.
7773 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7775 PR tree-optimization/83510
7776 * domwalk.c (set_all_edges_as_executable): New function.
7777 (dom_walker::dom_walker): Convert bool param
7778 "skip_unreachable_blocks" to enum reachability. Move setup of
7779 edge flags to set_all_edges_as_executable and only do it when
7780 reachability is REACHABLE_BLOCKS.
7781 * domwalk.h (enum dom_walker::reachability): New enum.
7782 (dom_walker::dom_walker): Convert bool param
7783 "skip_unreachable_blocks" to enum reachability.
7784 (set_all_edges_as_executable): New decl.
7785 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7786 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7787 "reachability".
7788 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7789 but converting true to REACHABLE_BLOCKS.
7790 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7791 * tree-vrp.c
7792 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7793 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7794 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7795 REACHABLE_BLOCKS.
7796 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7797 if check_all_array_refs will be called.
7799 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7801 * tree.c (selftest::test_location_wrappers): Add more test
7802 coverage.
7804 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7806 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7807 (selftest::test_bit_in_range): Likewise.
7809 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7811 PR testsuite/83888
7812 * doc/sourcebuild.texi (vect_float): Say that the selector
7813 only describes the situation when -funsafe-math-optimizations is on.
7814 (vect_float_strict): Document.
7816 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7818 PR tree-optimization/83965
7819 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7820 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7821 instead of checking only for a reduction.
7822 (vect_recog_widen_sum_pattern): Likewise.
7824 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7826 * predict.c (probably_never_executed): Only use precise profile info.
7827 (compute_function_frequency): Skip after inlining hack since we now
7828 have quality checking.
7830 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7832 * profile-count.h (profile_probability::very_unlikely,
7833 profile_probability::unlikely, profile_probability::even): Set
7834 precision to guessed.
7836 2018-01-23 Richard Biener <rguenther@suse.de>
7838 PR tree-optimization/83963
7839 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7840 Properly terminate dominator walk when crossing the exit edge not
7841 when visiting its source block.
7843 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7845 PR c++/83918
7846 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7847 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7849 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7851 PR tree-optimization/83957
7852 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7853 semicolon after for body surrounded by braces.
7855 PR tree-optimization/83081
7856 * profile-count.h (profile_probability::split): New method.
7857 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7858 Use profile_probability::split.
7859 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7860 when splitting a single conditional jump into 2.
7862 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7864 PR tree-optimization/69452
7865 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7866 decl.
7868 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7870 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7871 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7872 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7874 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7876 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7877 * config/rl78/rl78.md (movdi): New define_expand.
7878 * config/rl78/rl78.c (rl78_split_movdi): New function.
7880 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7882 PR target/83862
7883 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7884 no longer used.
7885 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7886 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7887 128-bit to produce an UNSPEC move to get the double word with the
7888 signbit and then a shift directly to do signbit.
7889 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7890 implementation with a new version that just does either a direct
7891 move or a regular move. Move memory interface to separate insns.
7892 Move insns so they are next to the expander.
7893 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7894 with signbit move. Split big and little endian case.
7895 (signbit<mode>2_dm_mem_le): Likewise.
7896 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7897 (signbit<mode>2_dm2): Likewise.
7899 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7901 * config/rl78/rl78.md (anddi3): New define_expand.
7903 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7905 * config/rl78/rl78.md (umindi3): New define_expand.
7907 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7909 * config/rl78/rl78.md (smindi3): New define_expand.
7911 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7913 * config/rl78/rl78.md (smaxdi3): New define_expand.
7915 2018-01-22 Carl Love <cel@us.ibm.com>
7917 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7918 LVX_V1TI): Add macro expansion.
7919 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7920 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7921 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7922 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7923 Change check to determine if the instruction is a byte reversing
7924 entry. Fix typo in comment.
7925 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7926 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7927 Add def_builtin calls for new builtins.
7928 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7929 Add define_insn expansion.
7931 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7933 * config/rl78/rl78.md (umaxdi3): New define_expand.
7935 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7937 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7938 for non-QImode registers.
7940 2018-01-22 Richard Biener <rguenther@suse.de>
7942 PR tree-optimization/83963
7943 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7944 including the loop exit block.
7945 (scop_detection::merge_sese): Likewise.
7946 (scop_detection::add_scop): Do it here instead.
7948 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7950 * doc/sourcebuild.texi (arm_softfloat): Document.
7952 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7954 PR gcc/77734
7955 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7956 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7957 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7959 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7960 David Edelsohn <dje.gcc@gmail.com>
7962 PR target/83946
7963 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7964 Change "crset eq" to "crset 2".
7965 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7966 (*call_indirect_aix<mode>_nospec): Likewise.
7967 (*call_value_indirect_aix<mode>_nospec): Likewise.
7968 (*call_indirect_elfv2<mode>_nospec): Likewise.
7969 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7970 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7971 change assembly output from . to $.
7972 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7973 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7974 (*tablejump<mode>_internal1_nospec): Likewise.
7976 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7978 PR target/80870
7979 * config/sh/sh_optimize_sett_clrt.cc:
7980 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7982 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7984 PR tree-optimization/83940
7985 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7986 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7987 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7988 use it to pass back the definition type.
7989 (vect_check_store_rhs): Likewise rhs_dt_out.
7990 (vect_build_gather_load_calls): Add a mask_dt argument and use
7991 it instead of a call to vect_is_simple_use.
7992 (vectorizable_store): Update calls to vect_check_load_store_mask
7993 and vect_check_store_rhs. Use the dt returned by the latter instead
7994 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7995 instead of calls to vect_is_simple_use. Pass the scalar rather
7996 than the vector operand to vect_is_simple_use when handling
7997 second and subsequent copies of an rhs value.
7998 (vectorizable_load): Update calls to vect_check_load_store_mask
7999 and vect_build_gather_load_calls. Use the cached mask_dt and
8000 gs_info.offset_dt instead of calls to vect_is_simple_use.
8002 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8004 PR middle-end/83945
8005 * tree-emutls.c: Include gimplify.h.
8006 (lower_emutls_2): New function.
8007 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8008 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8009 it before further processing.
8011 PR target/83930
8012 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8013 UINTVAL (trueop1) instead of INTVAL (op1).
8015 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8017 PR debug/81570
8018 PR debug/83728
8019 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8020 INCOMING_FRAME_SP_OFFSET if not defined.
8021 (scan_trace): Add ENTRY argument. If true and
8022 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8023 emit a note to adjust the CFA offset.
8024 (create_cfi_notes): Adjust scan_trace callers.
8025 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8026 INCOMING_FRAME_SP_OFFSET in the CIE.
8027 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8028 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8029 Likewise.
8030 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8031 * doc/tm.texi: Regenerated.
8033 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8035 PR rtl-optimization/83147
8036 * lra-constraints.c (remove_inheritance_pseudos): Use
8037 lra_substitute_pseudo_within_insn.
8039 2018-01-19 Tom de Vries <tom@codesourcery.com>
8040 Cesar Philippidis <cesar@codesourcery.com>
8042 PR target/83920
8043 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8045 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8047 PR target/83790
8048 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8049 spaces for function labels.
8051 2018-01-19 Martin Liska <mliska@suse.cz>
8053 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8054 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8055 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8056 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8057 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8058 (PRED_CONST_RETURN): Change from 69 to 65.
8059 (PRED_NULL_RETURN): Change from 91 to 71.
8060 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8061 (PRED_LOOP_GUARD): Change from 66 to 73.
8063 2018-01-19 Martin Liska <mliska@suse.cz>
8065 * predict.c (predict_insn_def): Add new assert.
8066 (struct branch_predictor): Change type to signed integer.
8067 (test_prediction_value_range): Amend test to cover
8068 PROB_UNINITIALIZED.
8069 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8070 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8071 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8072 (PRED_LOOP_IV_COMPARE): Likewise.
8073 * predict.h (PROB_UNINITIALIZED): Define new constant.
8075 2018-01-19 Martin Liska <mliska@suse.cz>
8077 * predict.c (dump_prediction): Add new format for
8078 analyze_brprob.py script which is enabled with -details
8079 suboption.
8080 * profile-count.h (precise_p): New function.
8082 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8084 PR tree-optimization/83922
8085 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8086 there are no statements that need masking.
8087 (vect_active_double_reduction_p): New function.
8088 (vect_analyze_loop_operations): Use it when handling phis that
8089 are not in the loop header.
8091 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8093 PR tree-optimization/83914
8094 * tree-vect-loop.c (vectorizable_induction): Don't convert
8095 init_expr or apply the peeling adjustment for inductions
8096 that are nested within the vectorized loop.
8098 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8100 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8101 instead of NEG.
8103 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8105 PR sanitizer/81715
8106 PR testsuite/83882
8107 * function.h (gimplify_parameters): Add gimple_seq * argument.
8108 * function.c: Include gimple.h and options.h.
8109 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8110 for the added local temporaries if needed.
8111 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8112 if there are any parameter cleanups, wrap whole body into a
8113 try/finally with the cleanups.
8115 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8117 PR target/82964
8118 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8119 Use GET_MODE_CLASS for scalar floating point.
8121 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8123 PR ipa/82256
8124 patch by PaX Team
8125 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8126 Fix call of call_cgraph_insertion_hooks.
8128 2018-01-18 Martin Sebor <msebor@redhat.com>
8130 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8132 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8134 PR ipa/83619
8135 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8136 frequencies.
8138 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8140 PR other/70268
8141 * common.opt: (-ffile-prefix-map): New option.
8142 * opts.c (common_handle_option): Defer it.
8143 * opts-global.c (handle_common_deferred_options): Handle it.
8144 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8145 * file-prefix-map.h: New file.
8146 (remap_debug_filename, add_debug_prefix_map): ...here.
8147 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8148 * final.c (debug_prefix_map, add_debug_prefix_map
8149 remap_debug_filename): Move to...
8150 * file-prefix-map.c: New file.
8151 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8152 generalize, get rid of alloca(), use strrchr() instead of strchr().
8153 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8154 Implement in terms of add_prefix_map().
8155 (remap_macro_filename, remap_debug_filename): Implement in term of
8156 remap_filename().
8157 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8158 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8159 * dbxout.c: Include file-prefix-map.h.
8160 * varasm.c: Likewise.
8161 * vmsdbgout.c: Likewise.
8162 * xcoffout.c: Likewise.
8163 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8164 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8165 * doc/invoke.texi (-ffile-prefix-map): Document.
8166 (-fdebug-prefix-map): Update description.
8168 2018-01-18 Martin Liska <mliska@suse.cz>
8170 * config/i386/i386.c (indirect_thunk_name): Document that also
8171 lfence is emitted.
8172 (output_indirect_thunk): Document why both instructions
8173 (pause and lfence) are generated.
8175 2018-01-18 Richard Biener <rguenther@suse.de>
8177 PR tree-optimization/83887
8178 * graphite-scop-detection.c
8179 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8180 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8181 (scop_detection::merge_sese): Re-implement with a flood-fill
8182 algorithm that properly finds a SESE region if it exists.
8184 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8186 PR c/61240
8187 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8188 pointer_diff optimizations use view_convert instead of convert.
8190 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8192 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8193 Generate different code for -mno-speculate-indirect-jumps.
8194 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8195 (*call_indirect_aix<mode>): Disable for
8196 -mno-speculate-indirect-jumps.
8197 (*call_indirect_aix<mode>_nospec): New define_insn.
8198 (*call_value_indirect_aix<mode>): Disable for
8199 -mno-speculate-indirect-jumps.
8200 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8201 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8202 -mno-speculate-indirect-jumps.
8203 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8205 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8207 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8208 long double type, set the flags for noting the default long double
8209 type, even if we don't pass or return a long double type.
8211 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8213 PR ipa/83051
8214 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8215 failure.
8217 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8219 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8220 support for merge[hl].
8221 (fold_mergehl_helper): New helper function.
8222 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8223 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8224 (altivec_vmrglw_direct): Add xxmrglw insn.
8226 2018-01-17 Andrew Waterman <andrew@sifive.com>
8228 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8229 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8231 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8233 PR lto/83121
8234 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8235 call the lto_location_cache before reading the
8236 DECL_SOURCE_LOCATION of the types.
8238 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8239 Richard Sandiford <richard.sandiford@linaro.org>
8241 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8242 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8243 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8244 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8245 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8246 Add declaration.
8247 * config/aarch64/constraints.md (aarch64_movti_operand):
8248 Limit immediates.
8249 * config/aarch64/predicates.md (Uti): Add new constraint.
8251 2018-01-17 Carl Love <cel@us.ibm.com>
8253 * config/rs6000/vsx.md (define_expand xl_len_r,
8254 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8255 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8256 lxvll.
8257 (define_expand, define_insn): Move the shift left from the
8258 define_insn to the define_expand for lxvl and stxvl instructions.
8259 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8260 and XL_LEN_R definitions to PURE.
8262 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8264 * config/i386/i386.c (indirect_thunk_name): Declare regno
8265 as unsigned int. Compare regno with INVALID_REGNUM.
8266 (output_indirect_thunk): Ditto.
8267 (output_indirect_thunk_function): Ditto.
8268 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8269 in the call to output_indirect_thunk_function.
8271 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8273 PR middle-end/83884
8274 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8275 rather than the size of inner_type to determine the stack slot size
8276 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8278 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8280 PR target/83546
8281 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8282 to PTA_SILVERMONT.
8284 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8286 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8287 endian Linux systems to optionally enable multilibs for selecting
8288 the long double type if the user configured an explicit type.
8289 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8290 have no long double multilibs if not defined.
8291 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8292 warn if the user used -mabi={ieee,ibm}longdouble and we built
8293 multilibs for long double.
8294 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8295 appropriate multilib option.
8296 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8297 multilib options.
8298 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8299 for building long double multilibs.
8300 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8302 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8304 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8305 copies.
8307 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8308 64 bits.
8309 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8310 128 bits.
8312 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8313 variables.
8315 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8316 return value.
8318 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8320 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8321 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8323 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8325 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8326 different rtl trees depending on TARGET_64BIT.
8327 (rs6000_gen_lvx): Likewise.
8329 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8331 * config/visium/visium.md (nop): Tweak comment.
8332 (hazard_nop): Likewise.
8334 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8336 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8337 -mspeculate-indirect-jumps.
8338 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8339 for -mno-speculate-indirect-jumps.
8340 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8341 (*call_value_indirect_elfv2<mode>): Disable for
8342 -mno-speculate-indirect-jumps.
8343 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8344 (indirect_jump): Emit different RTL for
8345 -mno-speculate-indirect-jumps.
8346 (*indirect_jump<mode>): Disable for
8347 -mno-speculate-indirect-jumps.
8348 (*indirect_jump<mode>_nospec): New define_insn.
8349 (tablejump): Emit different RTL for
8350 -mno-speculate-indirect-jumps.
8351 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8352 (tablejumpsi_nospec): New define_expand.
8353 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8354 (tablejumpdi_nospec): New define_expand.
8355 (*tablejump<mode>_internal1): Disable for
8356 -mno-speculate-indirect-jumps.
8357 (*tablejump<mode>_internal1_nospec): New define_insn.
8358 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8359 option.
8361 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8363 * caller-save.c (insert_save): Drop unnecessary parameter. All
8364 callers updated.
8366 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8367 Richard Biener <rguenth@suse.de>
8369 PR libgomp/83590
8370 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8371 return early, inline manually is_gimple_sizepos. Make sure if we
8372 call gimplify_expr we don't end up with a gimple constant.
8373 * tree.c (variably_modified_type_p): Don't return true for
8374 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8375 * gimplify.h (is_gimple_sizepos): Remove.
8377 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8379 PR tree-optimization/83857
8380 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8381 vectorizable_live_operation for pure SLP statements.
8382 (vectorizable_live_operation): Handle PHIs.
8384 2018-01-16 Richard Biener <rguenther@suse.de>
8386 PR tree-optimization/83867
8387 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8388 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8390 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8392 PR c/83844
8393 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8394 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8395 If off is not INTEGER_CST, issue a may not be aligned warning
8396 rather than isn't aligned. Use isn%'t rather than isn't.
8397 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8398 into MULT_EXPR.
8399 <case MULT_EXPR>: Improve the case when bottom and one of the
8400 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8401 operand, in that case check if the other operand is multiple of
8402 bottom divided by the INTEGER_CST operand.
8404 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8406 PR target/83858
8407 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8408 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8409 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8410 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8411 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8412 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8413 (pa_function_arg_size): New function.
8415 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8417 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8418 in a separate statement.
8420 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8422 PR tree-optimization/83847
8423 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8424 group gathers and scatters.
8426 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8428 PR rtl-optimization/86620
8429 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8431 PR rtl-optimization/83213
8432 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8433 to last if both are JUMP_INSNs.
8435 PR tree-optimization/83843
8436 * gimple-ssa-store-merging.c
8437 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8438 store_immediate_info for bswap/nop orig_stores.
8440 2018-01-15 Andrew Waterman <andrew@sifive.com>
8442 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8443 !TARGET_MUL.
8444 <UDIV>: Increase cost if !TARGET_DIV.
8446 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8448 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8449 (define_attr "cr_logical_3op"): New.
8450 (cceq_ior_compare): Adjust.
8451 (cceq_ior_compare_complement): Adjust.
8452 (*cceq_rev_compare): Adjust.
8453 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8454 (is_cracked_insn): Adjust.
8455 (insn_must_be_first_in_group): Adjust.
8456 * config/rs6000/40x.md: Adjust.
8457 * config/rs6000/440.md: Adjust.
8458 * config/rs6000/476.md: Adjust.
8459 * config/rs6000/601.md: Adjust.
8460 * config/rs6000/603.md: Adjust.
8461 * config/rs6000/6xx.md: Adjust.
8462 * config/rs6000/7450.md: Adjust.
8463 * config/rs6000/7xx.md: Adjust.
8464 * config/rs6000/8540.md: Adjust.
8465 * config/rs6000/cell.md: Adjust.
8466 * config/rs6000/e300c2c3.md: Adjust.
8467 * config/rs6000/e500mc.md: Adjust.
8468 * config/rs6000/e500mc64.md: Adjust.
8469 * config/rs6000/e5500.md: Adjust.
8470 * config/rs6000/e6500.md: Adjust.
8471 * config/rs6000/mpc.md: Adjust.
8472 * config/rs6000/power4.md: Adjust.
8473 * config/rs6000/power5.md: Adjust.
8474 * config/rs6000/power6.md: Adjust.
8475 * config/rs6000/power7.md: Adjust.
8476 * config/rs6000/power8.md: Adjust.
8477 * config/rs6000/power9.md: Adjust.
8478 * config/rs6000/rs64.md: Adjust.
8479 * config/rs6000/titan.md: Adjust.
8481 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8483 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8484 ix86_indirect_branch_register logic.
8486 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8488 * config/i386/constraints.md (Bs): Update
8489 ix86_indirect_branch_register check. Don't check
8490 ix86_indirect_branch_register with GOT_memory_operand.
8491 (Bw): Likewise.
8492 * config/i386/predicates.md (GOT_memory_operand): Don't check
8493 ix86_indirect_branch_register here.
8494 (GOT32_symbol_operand): Likewise.
8496 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8498 * config/i386/predicates.md (constant_call_address_operand):
8499 Rewrite ix86_indirect_branch_register logic.
8500 (sibcall_insn_operand): Likewise.
8502 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8504 * config/i386/constraints.md (Bs): Replace
8505 ix86_indirect_branch_thunk_register with
8506 ix86_indirect_branch_register.
8507 (Bw): Likewise.
8508 * config/i386/i386.md (indirect_jump): Likewise.
8509 (tablejump): Likewise.
8510 (*sibcall_memory): Likewise.
8511 (*sibcall_value_memory): Likewise.
8512 Peepholes of indirect call and jump via memory: Likewise.
8513 * config/i386/i386.opt: Likewise.
8514 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8515 (GOT_memory_operand): Likewise.
8516 (call_insn_operand): Likewise.
8517 (sibcall_insn_operand): Likewise.
8518 (GOT32_symbol_operand): Likewise.
8520 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8522 PR middle-end/83837
8523 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8524 type rather than type addr's type points to.
8525 (expand_omp_atomic_mutex): Likewise.
8526 (expand_omp_atomic): Likewise.
8528 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8530 PR target/83839
8531 * config/i386/i386.c (output_indirect_thunk_function): Use
8532 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8533 for __x86_return_thunk.
8535 2018-01-15 Richard Biener <rguenther@suse.de>
8537 PR middle-end/83850
8538 * expmed.c (extract_bit_field_1): Fix typo.
8540 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8542 PR target/83687
8543 * config/arm/iterators.md (VF): New mode iterator.
8544 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8545 Remove integer-related logic from pattern.
8546 (neon_vabd<mode>_3): Likewise.
8548 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8550 PR middle-end/82694
8551 * common.opt (fstrict-overflow): No longer an alias.
8552 (fwrapv-pointer): New option.
8553 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8554 also for pointer types based on flag_wrapv_pointer.
8555 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8556 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8557 opts->x_flag_wrapv got set.
8558 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8559 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8560 POINTER_TYPE_OVERFLOW_UNDEFINED.
8561 * match.pd: Likewise in address comparison pattern.
8562 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8564 2018-01-15 Richard Biener <rguenther@suse.de>
8566 PR lto/83804
8567 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8568 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8569 Reset type names to their identifier if their TYPE_DECL doesn't
8570 have linkage (and thus is used for ODR and devirt).
8571 (save_debug_info_for_decl): Remove.
8572 (save_debug_info_for_type): Likewise.
8573 (add_tree_to_fld_list): Adjust.
8574 * tree-pretty-print.c (dump_generic_node): Make dumping of
8575 type names more robust.
8577 2018-01-15 Richard Biener <rguenther@suse.de>
8579 * BASE-VER: Bump to 8.0.1.
8581 2018-01-14 Martin Sebor <msebor@redhat.com>
8583 PR other/83508
8584 * builtins.c (check_access): Avoid warning when the no-warning bit
8585 is set.
8587 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8589 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8590 * ira-color (allocno_hard_regs_compare): Likewise.
8592 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8594 PR target/83013
8595 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8596 Use .pushsection/.popsection.
8598 2018-01-14 Martin Sebor <msebor@redhat.com>
8600 PR c++/81327
8601 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8603 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8605 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8606 entry from extra_headers.
8607 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8608 extra_headers, make the list bitwise identical to the i?86-*-* one.
8610 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8612 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8613 -mcmodel=large with -mindirect-branch=thunk,
8614 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8615 -mfunction-return=thunk-extern.
8616 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8617 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8618 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8620 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8622 * config/i386/i386.c (print_reg): Print the name of the full
8623 integer register without '%'.
8624 (ix86_print_operand): Handle 'V'.
8625 * doc/extend.texi: Document 'V' modifier.
8627 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8629 * config/i386/constraints.md (Bs): Disallow memory operand for
8630 -mindirect-branch-register.
8631 (Bw): Likewise.
8632 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8633 (GOT_memory_operand): Likewise.
8634 (call_insn_operand): Likewise.
8635 (sibcall_insn_operand): Likewise.
8636 (GOT32_symbol_operand): Likewise.
8637 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8638 for -mindirect-branch-register.
8639 (tablejump): Likewise.
8640 (*sibcall_memory): Likewise.
8641 (*sibcall_value_memory): Likewise.
8642 Disallow peepholes of indirect call and jump via memory for
8643 -mindirect-branch-register.
8644 (*call_pop): Replace m with Bw.
8645 (*call_value_pop): Likewise.
8646 (*sibcall_pop_memory): Replace m with Bs.
8647 * config/i386/i386.opt (mindirect-branch-register): New option.
8648 * doc/invoke.texi: Document -mindirect-branch-register option.
8650 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8652 * config/i386/i386-protos.h (ix86_output_function_return): New.
8653 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8654 set function_return_type.
8655 (indirect_thunk_name): Add ret_p to indicate thunk for function
8656 return.
8657 (output_indirect_thunk_function): Pass false to
8658 indirect_thunk_name.
8659 (ix86_output_indirect_branch_via_reg): Likewise.
8660 (ix86_output_indirect_branch_via_push): Likewise.
8661 (output_indirect_thunk_function): Create alias for function
8662 return thunk if regno < 0.
8663 (ix86_output_function_return): New function.
8664 (ix86_handle_fndecl_attribute): Handle function_return.
8665 (ix86_attribute_table): Add function_return.
8666 * config/i386/i386.h (machine_function): Add
8667 function_return_type.
8668 * config/i386/i386.md (simple_return_internal): Use
8669 ix86_output_function_return.
8670 (simple_return_internal_long): Likewise.
8671 * config/i386/i386.opt (mfunction-return=): New option.
8672 (indirect_branch): Mention -mfunction-return=.
8673 * doc/extend.texi: Document function_return function attribute.
8674 * doc/invoke.texi: Document -mfunction-return= option.
8676 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8678 * config/i386/i386-opts.h (indirect_branch): New.
8679 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8680 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8681 with local indirect jump when converting indirect call and jump.
8682 (ix86_set_indirect_branch_type): New.
8683 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8684 (indirectlabelno): New.
8685 (indirect_thunk_needed): Likewise.
8686 (indirect_thunk_bnd_needed): Likewise.
8687 (indirect_thunks_used): Likewise.
8688 (indirect_thunks_bnd_used): Likewise.
8689 (INDIRECT_LABEL): Likewise.
8690 (indirect_thunk_name): Likewise.
8691 (output_indirect_thunk): Likewise.
8692 (output_indirect_thunk_function): Likewise.
8693 (ix86_output_indirect_branch_via_reg): Likewise.
8694 (ix86_output_indirect_branch_via_push): Likewise.
8695 (ix86_output_indirect_branch): Likewise.
8696 (ix86_output_indirect_jmp): Likewise.
8697 (ix86_code_end): Call output_indirect_thunk_function if needed.
8698 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8699 needed.
8700 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8701 (ix86_attribute_table): Add indirect_branch.
8702 * config/i386/i386.h (machine_function): Add indirect_branch_type
8703 and has_local_indirect_jump.
8704 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8705 to true.
8706 (tablejump): Likewise.
8707 (*indirect_jump): Use ix86_output_indirect_jmp.
8708 (*tablejump_1): Likewise.
8709 (simple_return_indirect_internal): Likewise.
8710 * config/i386/i386.opt (mindirect-branch=): New option.
8711 (indirect_branch): New.
8712 (keep): Likewise.
8713 (thunk): Likewise.
8714 (thunk-inline): Likewise.
8715 (thunk-extern): Likewise.
8716 * doc/extend.texi: Document indirect_branch function attribute.
8717 * doc/invoke.texi: Document -mindirect-branch= option.
8719 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8721 PR ipa/83051
8722 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8724 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8726 * ipa-inline.c (want_inline_small_function_p): Return false if
8727 inlining has already failed with CIF_FINAL_ERROR.
8728 (update_caller_keys): Call want_inline_small_function_p before
8729 can_inline_edge_p.
8730 (update_callee_keys): Likewise.
8732 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8734 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8735 New function.
8736 (rs6000_quadword_masked_address_p): Likewise.
8737 (quad_aligned_load_p): Likewise.
8738 (quad_aligned_store_p): Likewise.
8739 (const_load_sequence_p): Add comment to describe the outer-most loop.
8740 (mimic_memory_attributes_and_flags): New function.
8741 (rs6000_gen_stvx): Likewise.
8742 (replace_swapped_aligned_store): Likewise.
8743 (rs6000_gen_lvx): Likewise.
8744 (replace_swapped_aligned_load): Likewise.
8745 (replace_swapped_load_constant): Capitalize argument name in
8746 comment describing this function.
8747 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8748 and stores that access quad-word aligned addresses and replace
8749 with stvx or lvx instructions when appropriate.
8750 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8751 New function prototype.
8752 (rs6000_quadword_masked_address_p): Likewise.
8753 (rs6000_gen_lvx): Likewise.
8754 (rs6000_gen_stvx): Likewise.
8755 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8756 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8757 when memory address is aligned.
8758 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8759 this split to select lvx instruction when memory address is aligned.
8760 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8761 instruction when memory address is aligned.
8762 (*vsx_le_perm_load_v16qi): Likewise.
8763 (four unnamed splitters): Modify to select the stvx instruction
8764 when memory is aligned.
8766 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8768 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8769 which appears in the queue multiple times.
8771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8772 Alan Hayward <alan.hayward@arm.com>
8773 David Sherwood <david.sherwood@arm.com>
8775 * tree-vectorizer.h (vec_lower_bound): New structure.
8776 (_loop_vec_info): Add check_nonzero and lower_bounds.
8777 (LOOP_VINFO_CHECK_NONZERO): New macro.
8778 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8779 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8780 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8781 fields. Make seg_len the distance travelled, not including the
8782 access size.
8783 (dr_direction_indicator): Declare.
8784 (dr_zero_step_indicator): Likewise.
8785 (dr_known_forward_stride_p): Likewise.
8786 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8787 tree-ssanames.h.
8788 (runtime_alias_check_p): Allow runtime alias checks with
8789 variable strides.
8790 (operator ==): Compare access_size and align.
8791 (prune_runtime_alias_test_list): Rework for new distinction between
8792 the access_size and seg_len.
8793 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8794 segment lengths.
8795 (get_segment_min_max): New function.
8796 (create_intersect_range_checks): Use it.
8797 (dr_step_indicator): New function.
8798 (dr_direction_indicator): Likewise.
8799 (dr_zero_step_indicator): Likewise.
8800 (dr_known_forward_stride_p): Likewise.
8801 * tree-loop-distribution.c (data_ref_segment_size): Return
8802 DR_STEP * (niters - 1).
8803 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8804 constructor.
8805 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8806 (vect_preserves_scalar_order_p): New function, split out from...
8807 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8808 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8809 (vect_vfa_access_size): New function.
8810 (vect_vfa_align): Likewise.
8811 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8812 (dump_lower_bound): New function.
8813 (vect_check_lower_bound): Likewise.
8814 (vect_small_gap_p): Likewise.
8815 (vectorizable_with_step_bound_p): Likewise.
8816 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8817 depencies if the vectorization factor is 1. Convert the checks
8818 for nonzero steps into checks on the bounds of DR_STEP. Try using
8819 a bunds check for variable steps if the minimum required step is
8820 relatively small. Update calls to the dr_with_seg_len
8821 constructor and to vect_compile_time_alias.
8822 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8823 function.
8824 (vect_loop_versioning): Call it.
8825 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8826 when retrying.
8827 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8829 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8830 Alan Hayward <alan.hayward@arm.com>
8831 David Sherwood <david.sherwood@arm.com>
8833 * doc/sourcebuild.texi (vect_scatter_store): Document.
8834 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8835 optabs.
8836 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8837 Document.
8838 * genopinit.c (main): Add supports_vec_scatter_store and
8839 supports_vec_scatter_store_cached to target_optabs.
8840 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8841 IFN_MASK_SCATTER_STORE.
8842 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8843 functions.
8844 * internal-fn.h (internal_store_fn_p): Declare.
8845 (internal_fn_stored_value_index): Likewise.
8846 * internal-fn.c (scatter_store_direct): New macro.
8847 (expand_scatter_store_optab_fn): New function.
8848 (direct_scatter_store_optab_supported_p): New macro.
8849 (internal_store_fn_p): New function.
8850 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8851 IFN_MASK_SCATTER_STORE.
8852 (internal_fn_mask_index): Likewise.
8853 (internal_fn_stored_value_index): New function.
8854 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8855 for scatter stores.
8856 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8857 * optabs-query.c (supports_vec_scatter_store_p): New function.
8858 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8859 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8860 true for scatter stores.
8861 (vect_gather_scatter_fn_p): Handle scatter stores too.
8862 (vect_check_gather_scatter): Consider using scatter stores if
8863 supports_vec_scatter_store_p.
8864 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8865 scatter stores too.
8866 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8867 internal_fn_stored_value_index.
8868 (check_load_store_masking): Handle scatter stores too.
8869 (vect_get_store_rhs): Make public.
8870 (vectorizable_call): Use internal_store_fn_p.
8871 (vectorizable_store): Handle scatter store internal functions.
8872 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8873 when deciding whether the end of the group has been reached.
8874 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8875 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8876 (mask_scatter_store<mode>): New insns.
8878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8879 Alan Hayward <alan.hayward@arm.com>
8880 David Sherwood <david.sherwood@arm.com>
8882 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8883 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8884 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8885 function.
8886 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8887 Use vect_truncate_gather_scatter_offset if we can't treat the
8888 operation as a normal gather load or scatter store.
8889 (get_group_load_store_type): Take the gather_scatter_info
8890 as argument. Try using a gather load or scatter store for
8891 single-element groups.
8892 (get_load_store_type): Update calls to get_group_load_store_type
8893 and vect_use_strided_gather_scatters_p.
8895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8896 Alan Hayward <alan.hayward@arm.com>
8897 David Sherwood <david.sherwood@arm.com>
8899 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8900 optional tree argument.
8901 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8902 null target hooks.
8903 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8904 but continue to use the current value as a fallback.
8905 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8906 to compare the updates.
8907 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8908 (get_load_store_type): Use it when handling a strided access.
8909 (vect_get_strided_load_store_ops): New function.
8910 (vect_get_data_ptr_increment): Likewise.
8911 (vectorizable_load): Handle strided gather loads. Always pass
8912 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8914 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8915 Alan Hayward <alan.hayward@arm.com>
8916 David Sherwood <david.sherwood@arm.com>
8918 * doc/md.texi (gather_load@var{m}): Document.
8919 (mask_gather_load@var{m}): Likewise.
8920 * genopinit.c (main): Add supports_vec_gather_load and
8921 supports_vec_gather_load_cached to target_optabs.
8922 * optabs-tree.c (init_tree_optimization_optabs): Use
8923 ggc_cleared_alloc to allocate target_optabs.
8924 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8925 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8926 functions.
8927 * internal-fn.h (internal_load_fn_p): Declare.
8928 (internal_gather_scatter_fn_p): Likewise.
8929 (internal_fn_mask_index): Likewise.
8930 (internal_gather_scatter_fn_supported_p): Likewise.
8931 * internal-fn.c (gather_load_direct): New macro.
8932 (expand_gather_load_optab_fn): New function.
8933 (direct_gather_load_optab_supported_p): New macro.
8934 (direct_internal_fn_optab): New function.
8935 (internal_load_fn_p): Likewise.
8936 (internal_gather_scatter_fn_p): Likewise.
8937 (internal_fn_mask_index): Likewise.
8938 (internal_gather_scatter_fn_supported_p): Likewise.
8939 * optabs-query.c (supports_at_least_one_mode_p): New function.
8940 (supports_vec_gather_load_p): Likewise.
8941 * optabs-query.h (supports_vec_gather_load_p): Declare.
8942 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8943 and memory_type field.
8944 (NUM_PATTERNS): Bump to 15.
8945 * tree-vect-data-refs.c: Include internal-fn.h.
8946 (vect_gather_scatter_fn_p): New function.
8947 (vect_describe_gather_scatter_call): Likewise.
8948 (vect_check_gather_scatter): Try using internal functions for
8949 gather loads. Recognize existing calls to a gather load function.
8950 (vect_analyze_data_refs): Consider using gather loads if
8951 supports_vec_gather_load_p.
8952 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8953 (vect_get_gather_scatter_offset_type): Likewise.
8954 (vect_convert_mask_for_vectype): Likewise.
8955 (vect_add_conversion_to_patterm): Likewise.
8956 (vect_try_gather_scatter_pattern): Likewise.
8957 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8958 (vect_vect_recog_func_ptrs): Add it.
8959 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8960 internal_fn_mask_index and internal_gather_scatter_fn_p.
8961 (check_load_store_masking): Take the gather_scatter_info as an
8962 argument and handle gather loads.
8963 (vect_get_gather_scatter_ops): New function.
8964 (vectorizable_call): Check internal_load_fn_p.
8965 (vectorizable_load): Likewise. Handle gather load internal
8966 functions.
8967 (vectorizable_store): Update call to check_load_store_masking.
8968 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8969 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8970 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8971 (aarch64_gather_scale_operand_d): New predicates.
8972 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8973 (mask_gather_load<mode>): New insns.
8975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8976 Alan Hayward <alan.hayward@arm.com>
8977 David Sherwood <david.sherwood@arm.com>
8979 * optabs.def (fold_left_plus_optab): New optab.
8980 * doc/md.texi (fold_left_plus_@var{m}): Document.
8981 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8982 * internal-fn.c (fold_left_direct): Define.
8983 (expand_fold_left_optab_fn): Likewise.
8984 (direct_fold_left_optab_supported_p): Likewise.
8985 * fold-const-call.c (fold_const_fold_left): New function.
8986 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8987 * tree-parloops.c (valid_reduction_p): New function.
8988 (gather_scalar_reductions): Use it.
8989 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8990 (vect_finish_replace_stmt): Declare.
8991 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8992 (needs_fold_left_reduction_p): New function, split out from...
8993 (vect_is_simple_reduction): ...here. Accept reductions that
8994 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8995 (vect_force_simple_reduction): Also store the reduction type in
8996 the assignment's STMT_VINFO_REDUC_TYPE.
8997 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8998 (merge_with_identity): New function.
8999 (vect_expand_fold_left): Likewise.
9000 (vectorize_fold_left_reduction): Likewise.
9001 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9002 scalar phi in place for it. Check for target support and reject
9003 cases that would reassociate the operation. Defer the transform
9004 phase to vectorize_fold_left_reduction.
9005 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9006 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9007 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9011 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9012 call to ifc_temp_var.
9014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9015 Alan Hayward <alan.hayward@arm.com>
9016 David Sherwood <david.sherwood@arm.com>
9018 * target.def (legitimize_address_displacement): Take the original
9019 offset as a poly_int.
9020 * targhooks.h (default_legitimize_address_displacement): Update
9021 accordingly.
9022 * targhooks.c (default_legitimize_address_displacement): Likewise.
9023 * doc/tm.texi: Regenerate.
9024 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9025 as an argument, moving assert of ad->disp == ad->disp_term to...
9026 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9027 Try calling targetm.legitimize_address_displacement before expanding
9028 the address rather than afterwards, and adjust for the new interface.
9029 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9030 Match the new hook interface. Handle SVE addresses.
9031 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9032 new hook interface.
9034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9036 * Makefile.in (OBJS): Add early-remat.o.
9037 * target.def (select_early_remat_modes): New hook.
9038 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9039 * doc/tm.texi: Regenerate.
9040 * targhooks.h (default_select_early_remat_modes): Declare.
9041 * targhooks.c (default_select_early_remat_modes): New function.
9042 * timevar.def (TV_EARLY_REMAT): New timevar.
9043 * passes.def (pass_early_remat): New pass.
9044 * tree-pass.h (make_pass_early_remat): Declare.
9045 * early-remat.c: New file.
9046 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9047 function.
9048 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9051 Alan Hayward <alan.hayward@arm.com>
9052 David Sherwood <david.sherwood@arm.com>
9054 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9055 vfm1 with a bound_epilog parameter.
9056 (vect_do_peeling): Update calls accordingly, and move the prologue
9057 call earlier in the function. Treat the base bound_epilog as 0 for
9058 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9059 this base when peeling for gaps.
9060 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9061 with fully-masked loops.
9062 (vect_estimate_min_profitable_iters): Handle the single peeled
9063 iteration in that case.
9065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9066 Alan Hayward <alan.hayward@arm.com>
9067 David Sherwood <david.sherwood@arm.com>
9069 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9070 single-element interleaving even if the size is not a power of 2.
9071 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9072 accesses for single-element interleaving if the group size is
9073 not a power of 2.
9075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9076 Alan Hayward <alan.hayward@arm.com>
9077 David Sherwood <david.sherwood@arm.com>
9079 * doc/md.texi (fold_extract_last_@var{m}): Document.
9080 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9081 * optabs.def (fold_extract_last_optab): New optab.
9082 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9083 * internal-fn.c (fold_extract_direct): New macro.
9084 (expand_fold_extract_optab_fn): Likewise.
9085 (direct_fold_extract_optab_supported_p): Likewise.
9086 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9087 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9088 EXTRACT_LAST_REDUCTION.
9089 (get_initial_def_for_reduction): Do not create an initial vector
9090 for EXTRACT_LAST_REDUCTION reductions.
9091 (vectorizable_reduction): Leave the scalar phi in place for
9092 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9093 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9094 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9095 transform phase to vectorizable_condition.
9096 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9097 split out from...
9098 (vect_finish_stmt_generation): ...here.
9099 (vect_finish_replace_stmt): New function.
9100 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9101 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9102 pattern.
9103 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9106 Alan Hayward <alan.hayward@arm.com>
9107 David Sherwood <david.sherwood@arm.com>
9109 * doc/md.texi (extract_last_@var{m}): Document.
9110 * optabs.def (extract_last_optab): New optab.
9111 * internal-fn.def (EXTRACT_LAST): New internal function.
9112 * internal-fn.c (cond_unary_direct): New macro.
9113 (expand_cond_unary_optab_fn): Likewise.
9114 (direct_cond_unary_optab_supported_p): Likewise.
9115 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9116 loops using EXTRACT_LAST.
9117 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9118 (extract_last_<mode>): ...this optab.
9119 (vec_extract<mode><Vel>): Update accordingly.
9121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9122 Alan Hayward <alan.hayward@arm.com>
9123 David Sherwood <david.sherwood@arm.com>
9125 * target.def (empty_mask_is_expensive): New hook.
9126 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9127 * doc/tm.texi: Regenerate.
9128 * targhooks.h (default_empty_mask_is_expensive): Declare.
9129 * targhooks.c (default_empty_mask_is_expensive): New function.
9130 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9131 if the target says that empty masks are expensive.
9132 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9133 New function.
9134 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9137 Alan Hayward <alan.hayward@arm.com>
9138 David Sherwood <david.sherwood@arm.com>
9140 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9141 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9142 (vect_use_loop_mask_for_alignment_p): New function.
9143 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9144 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9145 niters_skip argument. Make sure that the first niters_skip elements
9146 of the first iteration are inactive.
9147 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9148 Update call to vect_set_loop_masks_directly.
9149 (get_misalign_in_elems): New function, split out from...
9150 (vect_gen_prolog_loop_niters): ...here.
9151 (vect_update_init_of_dr): Take a code argument that specifies whether
9152 the adjustment should be added or subtracted.
9153 (vect_update_init_of_drs): Likewise.
9154 (vect_prepare_for_masked_peels): New function.
9155 (vect_do_peeling): Skip prologue peeling if we're using a mask
9156 instead. Update call to vect_update_inits_of_drs.
9157 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9158 mask_skip_niters.
9159 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9160 alignment. Do not include the number of peeled iterations in
9161 the minimum threshold in that case.
9162 (vectorizable_induction): Adjust the start value down by
9163 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9164 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9165 Take the number of skipped iterations into account when calculating
9166 the loop bounds.
9167 * tree-vect-stmts.c (vect_gen_while_not): New function.
9169 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9170 Alan Hayward <alan.hayward@arm.com>
9171 David Sherwood <david.sherwood@arm.com>
9173 * doc/sourcebuild.texi (vect_fully_masked): Document.
9174 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9175 default value to 0.
9176 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9177 split out from...
9178 (vect_analyze_loop_2): ...here. Don't check the vectorization
9179 factor against the number of loop iterations if the loop is
9180 fully-masked.
9182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9183 Alan Hayward <alan.hayward@arm.com>
9184 David Sherwood <david.sherwood@arm.com>
9186 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9187 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9188 (dump_groups): Update accordingly.
9189 (iv_use::mem_type): New member variable.
9190 (address_p): New function.
9191 (record_use): Add a mem_type argument and initialize the new
9192 mem_type field.
9193 (record_group_use): Add a mem_type argument. Use address_p.
9194 Remove obsolete null checks of base_object. Update call to record_use.
9195 (find_interesting_uses_op): Update call to record_group_use.
9196 (find_interesting_uses_cond): Likewise.
9197 (find_interesting_uses_address): Likewise.
9198 (get_mem_type_for_internal_fn): New function.
9199 (find_address_like_use): Likewise.
9200 (find_interesting_uses_stmt): Try find_address_like_use before
9201 calling find_interesting_uses_op.
9202 (addr_offset_valid_p): Use the iv mem_type field as the type
9203 of the addressed memory.
9204 (add_autoinc_candidates): Likewise.
9205 (get_address_cost): Likewise.
9206 (split_small_address_groups_p): Use address_p.
9207 (split_address_groups): Likewise.
9208 (add_iv_candidate_for_use): Likewise.
9209 (autoinc_possible_for_pair): Likewise.
9210 (rewrite_groups): Likewise.
9211 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9212 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9213 (get_alias_ptr_type_for_ptr_address): New function.
9214 (rewrite_use_address): Rewrite address uses in calls that were
9215 identified by find_address_like_use.
9217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9218 Alan Hayward <alan.hayward@arm.com>
9219 David Sherwood <david.sherwood@arm.com>
9221 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9222 TARGET_MEM_REFs.
9223 * gimple-expr.h (is_gimple_addressable: Likewise.
9224 * gimple-expr.c (is_gimple_address): Likewise.
9225 * internal-fn.c (expand_call_mem_ref): New function.
9226 (expand_mask_load_optab_fn): Use it.
9227 (expand_mask_store_optab_fn): Likewise.
9229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9230 Alan Hayward <alan.hayward@arm.com>
9231 David Sherwood <david.sherwood@arm.com>
9233 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9234 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9235 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9236 (cond_umax@var{mode}): Document.
9237 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9238 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9239 (cond_umin_optab, cond_umax_optab): New optabs.
9240 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9241 (COND_IOR, COND_XOR): New internal functions.
9242 * internal-fn.h (get_conditional_internal_fn): Declare.
9243 * internal-fn.c (cond_binary_direct): New macro.
9244 (expand_cond_binary_optab_fn): Likewise.
9245 (direct_cond_binary_optab_supported_p): Likewise.
9246 (get_conditional_internal_fn): New function.
9247 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9248 Cope with reduction statements that are vectorized as calls rather
9249 than assignments.
9250 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9251 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9252 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9253 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9254 (UNSPEC_COND_EOR): New unspecs.
9255 (optab): Add mappings for them.
9256 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9257 (sve_int_op, sve_fp_op): New int attributes.
9259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9260 Alan Hayward <alan.hayward@arm.com>
9261 David Sherwood <david.sherwood@arm.com>
9263 * optabs.def (while_ult_optab): New optab.
9264 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9265 * internal-fn.def (WHILE_ULT): New internal function.
9266 * internal-fn.h (direct_internal_fn_supported_p): New override
9267 that takes two types as argument.
9268 * internal-fn.c (while_direct): New macro.
9269 (expand_while_optab_fn): New function.
9270 (convert_optab_supported_p): Likewise.
9271 (direct_while_optab_supported_p): New macro.
9272 * wide-int.h (wi::udiv_ceil): New function.
9273 * tree-vectorizer.h (rgroup_masks): New structure.
9274 (vec_loop_masks): New typedef.
9275 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9276 and fully_masked_p.
9277 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9278 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9279 (vect_max_vf): New function.
9280 (slpeel_make_loop_iterate_ntimes): Delete.
9281 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9282 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9283 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9284 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9285 internal-fn.h, stor-layout.h and optabs-query.h.
9286 (vect_set_loop_mask): New function.
9287 (add_preheader_seq): Likewise.
9288 (add_header_seq): Likewise.
9289 (interleave_supported_p): Likewise.
9290 (vect_maybe_permute_loop_masks): Likewise.
9291 (vect_set_loop_masks_directly): Likewise.
9292 (vect_set_loop_condition_masked): Likewise.
9293 (vect_set_loop_condition_unmasked): New function, split out from
9294 slpeel_make_loop_iterate_ntimes.
9295 (slpeel_make_loop_iterate_ntimes): Rename to..
9296 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9297 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9298 (vect_do_peeling): Update call accordingly.
9299 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9300 loops.
9301 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9302 mask_compare_type, can_fully_mask_p and fully_masked_p.
9303 (release_vec_loop_masks): New function.
9304 (_loop_vec_info): Use it to free the loop masks.
9305 (can_produce_all_loop_masks_p): New function.
9306 (vect_get_max_nscalars_per_iter): Likewise.
9307 (vect_verify_full_masking): Likewise.
9308 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9309 retries, and free the mask rgroups before retrying. Check loop-wide
9310 reasons for disallowing fully-masked loops. Make the final decision
9311 about whether use a fully-masked loop or not.
9312 (vect_estimate_min_profitable_iters): Do not assume that peeling
9313 for the number of iterations will be needed for fully-masked loops.
9314 (vectorizable_reduction): Disable fully-masked loops.
9315 (vectorizable_live_operation): Likewise.
9316 (vect_halve_mask_nunits): New function.
9317 (vect_double_mask_nunits): Likewise.
9318 (vect_record_loop_mask): Likewise.
9319 (vect_get_loop_mask): Likewise.
9320 (vect_transform_loop): Handle the case in which the final loop
9321 iteration might handle a partial vector. Call vect_set_loop_condition
9322 instead of slpeel_make_loop_iterate_ntimes.
9323 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9324 (check_load_store_masking): New function.
9325 (prepare_load_store_mask): Likewise.
9326 (vectorizable_store): Handle fully-masked loops.
9327 (vectorizable_load): Likewise.
9328 (supportable_widening_operation): Use vect_halve_mask_nunits for
9329 booleans.
9330 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9331 (vect_gen_while): New function.
9332 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9333 (aarch64_uqdec<mode>): New insn.
9335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9336 Alan Hayward <alan.hayward@arm.com>
9337 David Sherwood <david.sherwood@arm.com>
9339 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9340 (reduc_xor_scal_optab): New optabs.
9341 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9342 (reduc_xor_scal_@var{m}): Document.
9343 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9344 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9345 internal functions.
9346 * fold-const-call.c (fold_const_call): Handle them.
9347 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9348 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9349 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9350 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9351 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9352 (UNSPEC_XORV): New unspecs.
9353 (optab): Add entries for them.
9354 (BITWISEV): New int iterator.
9355 (bit_reduc_op): New int attributes.
9357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9358 Alan Hayward <alan.hayward@arm.com>
9359 David Sherwood <david.sherwood@arm.com>
9361 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9362 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9363 * optabs.def (vec_shl_insert_optab): New optab.
9364 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9365 (duplicate_and_interleave): Likewise.
9366 * tree-vect-loop.c: Include internal-fn.h.
9367 (neutral_op_for_slp_reduction): New function, split out from
9368 get_initial_defs_for_reduction.
9369 (get_initial_def_for_reduction): Handle option 2 for variable-length
9370 vectors by loading the neutral value into a vector and then shifting
9371 the initial value into element 0.
9372 (get_initial_defs_for_reduction): Replace the code argument with
9373 the neutral value calculated by neutral_op_for_slp_reduction.
9374 Use gimple_build_vector for constant-length vectors.
9375 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9376 but the first group_size elements have a neutral value.
9377 Use duplicate_and_interleave otherwise.
9378 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9379 Update call to get_initial_defs_for_reduction. Handle SLP
9380 reductions for variable-length vectors by creating one vector
9381 result for each scalar result, with the elements associated
9382 with other scalar results stubbed out with the neutral value.
9383 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9384 Require IFN_VEC_SHL_INSERT for double reductions on
9385 variable-length vectors, or SLP reductions that have
9386 a neutral value. Require can_duplicate_and_interleave_p
9387 support for variable-length unchained SLP reductions if there
9388 is no neutral value, such as for MIN/MAX reductions. Also require
9389 the number of vector elements to be a multiple of the number of
9390 SLP statements when doing variable-length unchained SLP reductions.
9391 Update call to vect_create_epilog_for_reduction.
9392 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9393 and remove initial values.
9394 (duplicate_and_interleave): Make public.
9395 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9396 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9399 Alan Hayward <alan.hayward@arm.com>
9400 David Sherwood <david.sherwood@arm.com>
9402 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9403 (can_duplicate_and_interleave_p): New function.
9404 (vect_get_and_check_slp_defs): Take the vector of statements
9405 rather than just the current one. Remove excess parentheses.
9406 Restriction rejectinon of vect_constant_def and vect_external_def
9407 for variable-length vectors to boolean types, or types for which
9408 can_duplicate_and_interleave_p is false.
9409 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9410 (duplicate_and_interleave): New function.
9411 (vect_get_constant_vectors): Use gimple_build_vector for
9412 constant-length vectors and suitable variable-length constant
9413 vectors. Use duplicate_and_interleave for other variable-length
9414 vectors. Don't defer the update when inserting new statements.
9416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9417 Alan Hayward <alan.hayward@arm.com>
9418 David Sherwood <david.sherwood@arm.com>
9420 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9421 min_profitable_iters doesn't go negative.
9423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9424 Alan Hayward <alan.hayward@arm.com>
9425 David Sherwood <david.sherwood@arm.com>
9427 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9428 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9429 * optabs.def (vec_mask_load_lanes_optab): New optab.
9430 (vec_mask_store_lanes_optab): Likewise.
9431 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9432 (MASK_STORE_LANES): Likewise.
9433 * internal-fn.c (mask_load_lanes_direct): New macro.
9434 (mask_store_lanes_direct): Likewise.
9435 (expand_mask_load_optab_fn): Handle masked operations.
9436 (expand_mask_load_lanes_optab_fn): New macro.
9437 (expand_mask_store_optab_fn): Handle masked operations.
9438 (expand_mask_store_lanes_optab_fn): New macro.
9439 (direct_mask_load_lanes_optab_supported_p): Likewise.
9440 (direct_mask_store_lanes_optab_supported_p): Likewise.
9441 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9442 parameter.
9443 (vect_load_lanes_supported): Likewise.
9444 * tree-vect-data-refs.c (strip_conversion): New function.
9445 (can_group_stmts_p): Likewise.
9446 (vect_analyze_data_ref_accesses): Use it instead of checking
9447 for a pair of assignments.
9448 (vect_store_lanes_supported): Take a masked_p parameter.
9449 (vect_load_lanes_supported): Likewise.
9450 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9451 vect_store_lanes_supported and vect_load_lanes_supported.
9452 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9453 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9454 parameter. Don't allow gaps for masked accesses.
9455 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9456 and vect_load_lanes_supported.
9457 (get_load_store_type): Take a masked_p parameter and update
9458 call to get_group_load_store_type.
9459 (vectorizable_store): Update call to get_load_store_type.
9460 Handle IFN_MASK_STORE_LANES.
9461 (vectorizable_load): Update call to get_load_store_type.
9462 Handle IFN_MASK_LOAD_LANES.
9464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9465 Alan Hayward <alan.hayward@arm.com>
9466 David Sherwood <david.sherwood@arm.com>
9468 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9469 modes for SVE.
9470 * config/aarch64/aarch64-protos.h
9471 (aarch64_sve_struct_memory_operand_p): Declare.
9472 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9473 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9474 (VPRED, vpred): Handle SVE structure modes.
9475 * config/aarch64/constraints.md (Utx): New constraint.
9476 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9477 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9478 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9479 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9480 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9481 structure modes. Split into pieces after RA.
9482 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9483 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9484 New patterns.
9485 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9486 SVE structure modes.
9487 (aarch64_classify_address): Likewise.
9488 (sizetochar): Move earlier in file.
9489 (aarch64_print_operand): Handle SVE register lists.
9490 (aarch64_array_mode): New function.
9491 (aarch64_sve_struct_memory_operand_p): Likewise.
9492 (TARGET_ARRAY_MODE): Redefine.
9494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9495 Alan Hayward <alan.hayward@arm.com>
9496 David Sherwood <david.sherwood@arm.com>
9498 * target.def (array_mode): New target hook.
9499 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9500 * doc/tm.texi: Regenerate.
9501 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9502 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9503 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9504 targetm.array_mode.
9505 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9506 type sizes.
9508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9509 Alan Hayward <alan.hayward@arm.com>
9510 David Sherwood <david.sherwood@arm.com>
9512 * fold-const.c (fold_binary_loc): Check the argument types
9513 rather than the result type when testing for a vector operation.
9515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9517 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9518 * doc/tm.texi: Regenerate.
9520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9521 Alan Hayward <alan.hayward@arm.com>
9522 David Sherwood <david.sherwood@arm.com>
9524 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9525 (sve): Document new AArch64 extension.
9526 * doc/md.texi (w): Extend the description of the AArch64
9527 constraint to include SVE vectors.
9528 (Upl, Upa): Document new AArch64 predicate constraints.
9529 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9530 enum.
9531 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9532 (msve-vector-bits=): New option.
9533 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9534 SVE when these are disabled.
9535 (sve): New extension.
9536 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9537 modes. Adjust their number of units based on aarch64_sve_vg.
9538 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9539 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9540 aarch64_addr_query_type.
9541 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9542 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9543 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9544 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9545 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9546 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9547 (aarch64_simd_imm_zero_p): Delete.
9548 (aarch64_check_zero_based_sve_index_immediate): Declare.
9549 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9550 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9551 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9552 (aarch64_sve_float_mul_immediate_p): Likewise.
9553 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9554 rather than an rtx.
9555 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9556 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9557 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9558 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9559 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9560 (aarch64_regmode_natural_size): Likewise.
9561 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9562 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9563 left one place.
9564 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9565 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9566 for VG and the SVE predicate registers.
9567 (V_ALIASES): Add a "z"-prefixed alias.
9568 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9569 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9570 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9571 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9572 (REG_CLASS_NAMES): Add entries for them.
9573 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9574 and the predicate registers.
9575 (aarch64_sve_vg): Declare.
9576 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9577 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9578 (REGMODE_NATURAL_SIZE): Define.
9579 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9580 SVE macros.
9581 * config/aarch64/aarch64.c: Include cfgrtl.h.
9582 (simd_immediate_info): Add a constructor for series vectors,
9583 and an associated step field.
9584 (aarch64_sve_vg): New variable.
9585 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9586 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9587 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9588 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9589 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9590 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9591 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9592 (aarch64_get_mask_mode): New functions.
9593 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9594 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9595 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9596 predicate modes and predicate registers. Explicitly restrict
9597 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9598 to store a vector mode if it is recognized by
9599 aarch64_classify_vector_mode.
9600 (aarch64_regmode_natural_size): New function.
9601 (aarch64_hard_regno_caller_save_mode): Return the original mode
9602 for predicates.
9603 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9604 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9605 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9606 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9607 functions.
9608 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9609 does not overlap dest if the function is frame-related. Handle
9610 SVE constants.
9611 (aarch64_split_add_offset): New function.
9612 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9613 them aarch64_add_offset.
9614 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9615 and update call to aarch64_sub_sp.
9616 (aarch64_add_cfa_expression): New function.
9617 (aarch64_expand_prologue): Pass extra temporary registers to the
9618 functions above. Handle the case in which we need to emit new
9619 DW_CFA_expressions for registers that were originally saved
9620 relative to the stack pointer, but now have to be expressed
9621 relative to the frame pointer.
9622 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9623 functions above.
9624 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9625 IP0 and IP1 values for SVE frames.
9626 (aarch64_expand_vec_series): New function.
9627 (aarch64_expand_sve_widened_duplicate): Likewise.
9628 (aarch64_expand_sve_const_vector): Likewise.
9629 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9630 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9631 into the register, rather than emitting a SET directly.
9632 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9633 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9634 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9635 (offset_9bit_signed_scaled_p): New functions.
9636 (aarch64_replicate_bitmask_imm): New function.
9637 (aarch64_bitmask_imm): Use it.
9638 (aarch64_cannot_force_const_mem): Reject expressions involving
9639 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9640 (aarch64_classify_index): Handle SVE indices, by requiring
9641 a plain register index with a scale that matches the element size.
9642 (aarch64_classify_address): Handle SVE addresses. Assert that
9643 the mode of the address is VOIDmode or an integer mode.
9644 Update call to aarch64_classify_symbol.
9645 (aarch64_classify_symbolic_expression): Update call to
9646 aarch64_classify_symbol.
9647 (aarch64_const_vec_all_in_range_p): New function.
9648 (aarch64_print_vector_float_operand): Likewise.
9649 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9650 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9651 and the FP immediates 1.0 and 0.5.
9652 (aarch64_print_address_internal): Handle SVE addresses.
9653 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9654 (aarch64_regno_regclass): Handle predicate registers.
9655 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9656 data modes.
9657 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9658 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9659 (aarch64_convert_sve_vector_bits): New function.
9660 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9661 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9662 rather than an rtx.
9663 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9664 Handle SVE vector and predicate modes. Accept VL-based constants
9665 that need only one temporary register, and VL offsets that require
9666 no temporary registers.
9667 (aarch64_conditional_register_usage): Mark the predicate registers
9668 as fixed if SVE isn't available.
9669 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9670 Return true for SVE vector and predicate modes.
9671 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9672 rather than an unsigned int. Handle SVE modes.
9673 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9674 SVE modes.
9675 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9676 if SVE is enabled.
9677 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9678 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9679 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9680 (aarch64_sve_float_mul_immediate_p): New functions.
9681 (aarch64_sve_valid_immediate): New function.
9682 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9683 Explicitly reject structure modes. Check for INDEX constants.
9684 Handle PTRUE and PFALSE constants.
9685 (aarch64_check_zero_based_sve_index_immediate): New function.
9686 (aarch64_simd_imm_zero_p): Delete.
9687 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9688 vector modes. Accept constants in the range of CNT[BHWD].
9689 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9690 ask for an Advanced SIMD mode.
9691 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9692 (aarch64_simd_vector_alignment): Handle SVE predicates.
9693 (aarch64_vectorize_preferred_vector_alignment): New function.
9694 (aarch64_simd_vector_alignment_reachable): Use it instead of
9695 the vector size.
9696 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9697 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9698 functions.
9699 (MAX_VECT_LEN): Delete.
9700 (expand_vec_perm_d): Add a vec_flags field.
9701 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9702 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9703 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9704 for SVE modes.
9705 (aarch64_evpc_rev): Rename to...
9706 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9707 (aarch64_evpc_rev_global): New function.
9708 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9709 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9710 MAX_VECT_LEN.
9711 (aarch64_evpc_sve_tbl): New function.
9712 (aarch64_expand_vec_perm_const_1): Update after rename of
9713 aarch64_evpc_rev. Handle SVE permutes too, trying
9714 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9715 than aarch64_evpc_tbl.
9716 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9717 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9718 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9719 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9720 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9721 (aarch64_expand_sve_vcond): New functions.
9722 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9723 of aarch64_vector_mode_p.
9724 (aarch64_dwarf_poly_indeterminate_value): New function.
9725 (aarch64_compute_pressure_classes): Likewise.
9726 (aarch64_can_change_mode_class): Likewise.
9727 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9728 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9729 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9730 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9731 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9732 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9733 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9734 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9735 constraints.
9736 (Dn, Dl, Dr): Accept const as well as const_vector.
9737 (Dz): Likewise. Compare against CONST0_RTX.
9738 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9739 of "vector" where appropriate.
9740 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9741 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9742 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9743 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9744 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9745 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9746 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9747 (v_int_equiv): Extend to SVE modes.
9748 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9749 mode attributes.
9750 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9751 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9752 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9753 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9754 (SVE_COND_FP_CMP): New int iterators.
9755 (perm_hilo): Handle the new unpack unspecs.
9756 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9757 attributes.
9758 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9759 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9760 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9761 (aarch64_equality_operator, aarch64_constant_vector_operand)
9762 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9763 (aarch64_sve_nonimmediate_operand): Likewise.
9764 (aarch64_sve_general_operand): Likewise.
9765 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9766 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9767 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9768 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9769 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9770 (aarch64_sve_float_arith_immediate): Likewise.
9771 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9772 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9773 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9774 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9775 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9776 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9777 (aarch64_sve_float_arith_operand): Likewise.
9778 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9779 (aarch64_sve_float_mul_operand): Likewise.
9780 (aarch64_sve_vec_perm_operand): Likewise.
9781 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9782 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9783 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9784 as well as const_vector.
9785 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9786 in file. Use CONST0_RTX and CONSTM1_RTX.
9787 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9788 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9789 Use aarch64_simd_imm_zero.
9790 * config/aarch64/aarch64-sve.md: New file.
9791 * config/aarch64/aarch64.md: Include it.
9792 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9793 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9794 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9795 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9796 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9797 (sve): New attribute.
9798 (enabled): Disable instructions with the sve attribute unless
9799 TARGET_SVE.
9800 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9801 aarch64_expand_mov_immediate.
9802 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9803 CNT[BHSD] immediates.
9804 (movti): Split CONST_POLY_INT moves into two halves.
9805 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9806 Split additions that need a temporary here if the destination
9807 is the stack pointer.
9808 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9809 (*add<mode>3_poly_1): New instruction.
9810 (set_clobber_cc): New expander.
9812 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9814 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9815 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9816 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9817 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9818 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9819 Change innermode from fixed_mode_size to machine_mode.
9820 (simplify_subreg): Update call accordingly. Handle a constant-sized
9821 subreg of a variable-length CONST_VECTOR.
9823 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9824 Alan Hayward <alan.hayward@arm.com>
9825 David Sherwood <david.sherwood@arm.com>
9827 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9828 (add_offset_to_base): New function, split out from...
9829 (create_mem_ref): ...here. When handling a scale other than 1,
9830 check first whether the address is valid without the offset.
9831 Add it into the base if so, leaving the index and scale as-is.
9833 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9835 PR c++/83778
9836 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9837 fold_for_warn before checking if arg2 is INTEGER_CST.
9839 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9841 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9842 (store_multiple_operation): Delete.
9843 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9844 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9845 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9846 guarded by TARGET_STRING.
9847 (rs6000_output_load_multiple): Delete.
9848 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9849 OPTION_MASK_STRING / TARGET_STRING handling.
9850 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9851 (const rs6000_opt_masks) <"string">: Change mask to 0.
9852 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9853 (MASK_STRING): Delete.
9854 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9855 parts. Simplify.
9856 (load_multiple): Delete.
9857 (*ldmsi8): Delete.
9858 (*ldmsi7): Delete.
9859 (*ldmsi6): Delete.
9860 (*ldmsi5): Delete.
9861 (*ldmsi4): Delete.
9862 (*ldmsi3): Delete.
9863 (store_multiple): Delete.
9864 (*stmsi8): Delete.
9865 (*stmsi7): Delete.
9866 (*stmsi6): Delete.
9867 (*stmsi5): Delete.
9868 (*stmsi4): Delete.
9869 (*stmsi3): Delete.
9870 (movmemsi_8reg): Delete.
9871 (corresponding unnamed define_insn): Delete.
9872 (movmemsi_6reg): Delete.
9873 (corresponding unnamed define_insn): Delete.
9874 (movmemsi_4reg): Delete.
9875 (corresponding unnamed define_insn): Delete.
9876 (movmemsi_2reg): Delete.
9877 (corresponding unnamed define_insn): Delete.
9878 (movmemsi_1reg): Delete.
9879 (corresponding unnamed define_insn): Delete.
9880 * config/rs6000/rs6000.opt (mno-string): New.
9881 (mstring): Replace by deprecation warning stub.
9882 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9884 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9886 * regrename.c (regrename_do_replace): If replacing the same
9887 reg multiple times, try to reuse last created gen_raw_REG.
9889 PR debug/81155
9890 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9891 main to workaround a bug in GDB.
9893 2018-01-12 Tom de Vries <tom@codesourcery.com>
9895 PR target/83737
9896 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9898 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9900 PR rtl-optimization/80481
9901 * ira-color.c (get_cap_member): New function.
9902 (allocnos_conflict_by_live_ranges_p): Use it.
9903 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9904 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9906 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9908 PR target/83628
9909 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9910 (*saddl_se_1): Ditto.
9911 (*ssubsi_1): Ditto.
9912 (*ssubl_se_1): Ditto.
9914 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9916 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9917 rather than wi::to_widest for DR_INITs.
9918 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9919 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9920 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9921 INTEGER_CSTs.
9922 (vect_analyze_group_access_1): Note that here.
9924 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9926 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9927 polynomial type sizes.
9929 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9931 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9932 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9933 (gimple_add_tmp_var): Likewise.
9935 2018-01-12 Martin Liska <mliska@suse.cz>
9937 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9938 (gimple_alloc_sizes): Likewise.
9939 (dump_gimple_statistics): Use PRIu64 in printf format.
9940 * gimple.h: Change uint64_t to int.
9942 2018-01-12 Martin Liska <mliska@suse.cz>
9944 * tree-core.h: Use uint64_t instead of int.
9945 * tree.c (tree_node_counts): Likewise.
9946 (tree_node_sizes): Likewise.
9947 (dump_tree_statistics): Use PRIu64 in printf format.
9949 2018-01-12 Martin Liska <mliska@suse.cz>
9951 * Makefile.in: As qsort_chk is implemented in vec.c, add
9952 vec.o to linkage of gencfn-macros.
9953 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9954 passing the info to record_node_allocation_statistics.
9955 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9956 and pass the info.
9957 * ggc-common.c (struct ggc_usage): Add operator== and use
9958 it in operator< and compare function.
9959 * mem-stats.h (struct mem_usage): Likewise.
9960 * vec.c (struct vec_usage): Remove operator< and compare
9961 function. Can be simply inherited.
9963 2018-01-12 Martin Jambor <mjambor@suse.cz>
9965 PR target/81616
9966 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9967 * tree-ssa-math-opts.c: Include domwalk.h.
9968 (convert_mult_to_fma_1): New function.
9969 (fma_transformation_info): New type.
9970 (fma_deferring_state): Likewise.
9971 (cancel_fma_deferring): New function.
9972 (result_of_phi): Likewise.
9973 (last_fma_candidate_feeds_initial_phi): Likewise.
9974 (convert_mult_to_fma): Added deferring logic, split actual
9975 transformation to convert_mult_to_fma_1.
9976 (math_opts_dom_walker): New type.
9977 (math_opts_dom_walker::after_dom_children): New method, body moved
9978 here from pass_optimize_widening_mul::execute, added deferring logic
9979 bits.
9980 (pass_optimize_widening_mul::execute): Moved most of code to
9981 math_opts_dom_walker::after_dom_children.
9982 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9983 * config/i386/i386.c (ix86_option_override_internal): Added
9984 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9986 2018-01-12 Richard Biener <rguenther@suse.de>
9988 PR debug/83157
9989 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9990 inline instance vars.
9992 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9994 PR target/81819
9995 * config/rx/rx.c (rx_is_restricted_memory_address):
9996 Handle SUBREG case.
9998 2018-01-12 Richard Biener <rguenther@suse.de>
10000 PR tree-optimization/80846
10001 * target.def (split_reduction): New target hook.
10002 * targhooks.c (default_split_reduction): New function.
10003 * targhooks.h (default_split_reduction): Declare.
10004 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10005 target requests first reduce vectors by combining low and high
10006 parts.
10007 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10008 (get_vectype_for_scalar_type_and_size): Export.
10009 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10010 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10011 * doc/tm.texi: Regenerate.
10012 * config/i386/i386.c (ix86_split_reduction): Implement
10013 TARGET_VECTORIZE_SPLIT_REDUCTION.
10015 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10017 PR target/83368
10018 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10019 in PIC mode except for TARGET_VXWORKS_RTP.
10020 * config/sparc/sparc.c: Include cfgrtl.h.
10021 (TARGET_INIT_PIC_REG): Define.
10022 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10023 (sparc_pic_register_p): New predicate.
10024 (sparc_legitimate_address_p): Use it.
10025 (sparc_legitimize_pic_address): Likewise.
10026 (sparc_delegitimize_address): Likewise.
10027 (sparc_mode_dependent_address_p): Likewise.
10028 (gen_load_pcrel_sym): Remove 4th parameter.
10029 (load_got_register): Adjust call to above. Remove obsolete stuff.
10030 (sparc_expand_prologue): Do not call load_got_register here.
10031 (sparc_flat_expand_prologue): Likewise.
10032 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10033 (sparc_use_pseudo_pic_reg): New function.
10034 (sparc_init_pic_reg): Likewise.
10035 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10036 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10038 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10040 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10041 Add item for branch_cost.
10043 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10045 PR rtl-optimization/83565
10046 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10047 not extend the result to a larger mode for rotate operations.
10048 (num_sign_bit_copies1): Likewise.
10050 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10052 PR target/40411
10053 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10054 -symbolic.
10055 Use values-Xc.o for -pedantic.
10056 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10058 2018-01-12 Martin Liska <mliska@suse.cz>
10060 PR ipa/83054
10061 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10062 New function.
10063 (possible_polymorphic_call_targets): Use it.
10064 (ipa_devirt): Likewise.
10066 2018-01-12 Martin Liska <mliska@suse.cz>
10068 * profile-count.h (enum profile_quality): Use 0 as invalid
10069 enum value of profile_quality.
10071 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10073 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10074 -mext-string options.
10076 2018-01-12 Richard Biener <rguenther@suse.de>
10078 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10079 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10080 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10081 Likewise.
10082 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10084 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10086 * configure.ac (--with-long-double-format): Add support for the
10087 configuration option to change the default long double format on
10088 PowerPC systems.
10089 * config.gcc (powerpc*-linux*-*): Likewise.
10090 * configure: Regenerate.
10091 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10092 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10093 used without modification.
10095 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10097 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10098 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10099 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10100 MISC_BUILTIN_SPEC_BARRIER.
10101 (rs6000_init_builtins): Likewise.
10102 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10103 enum value.
10104 (speculation_barrier): New define_insn.
10105 * doc/extend.texi: Document __builtin_speculation_barrier.
10107 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10109 PR target/83203
10110 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10111 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10112 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10113 iterators.
10114 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10115 integral modes instead of "ss" and "sd".
10116 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10117 vectors with 32-bit and 64-bit elements.
10118 (vecdupssescalarmodesuffix): New mode attribute.
10119 (vec_dup<mode>): Use it.
10121 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10123 PR target/83330
10124 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10125 frame if argument is passed on stack.
10127 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10129 PR target/82682
10130 * ree.c (combine_reaching_defs): Optimize also
10131 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10132 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10134 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10136 PR middle-end/83189
10137 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10139 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10141 PR middle-end/83718
10142 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10143 after they are computed.
10145 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10147 PR tree-optimization/83695
10148 * gimple-loop-linterchange.cc
10149 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10150 reset cached scev information after interchange.
10151 (pass_linterchange::execute): Remove call to scev_reset_htab.
10153 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10155 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10156 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10157 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10158 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10159 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10160 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10161 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10162 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10163 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10164 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10165 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10166 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10167 (V_lane_reg): Likewise.
10168 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10169 New define_expand.
10170 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10171 (vfmal_lane_low<mode>_intrinsic,
10172 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10173 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10174 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10175 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10176 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10177 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10179 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10181 * config/arm/arm-cpus.in (fp16fml): New feature.
10182 (ALL_SIMD): Add fp16fml.
10183 (armv8.2-a): Add fp16fml as an option.
10184 (armv8.3-a): Likewise.
10185 (armv8.4-a): Add fp16fml as part of fp16.
10186 * config/arm/arm.h (TARGET_FP16FML): Define.
10187 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10188 when appropriate.
10189 * config/arm/arm-modes.def (V2HF): Define.
10190 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10191 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10192 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10193 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10194 vfmsl_low, vfmsl_high): New set of builtins.
10195 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10196 (vfml_op): New code attribute.
10197 (VFMLHALVES): New int iterator.
10198 (VFML, VFMLSEL): New mode attributes.
10199 (V_reg): Define mapping for V2HF.
10200 (V_hi, V_lo): New mode attributes.
10201 (VF_constraint): Likewise.
10202 (vfml_half, vfml_half_selector): New int attributes.
10203 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10204 define_expand.
10205 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10206 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10207 New define_insn.
10208 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10209 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10210 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10211 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10212 documentation.
10213 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10214 Document new effective target and option set.
10216 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10218 * config/arm/arm-cpus.in (armv8_4): New feature.
10219 (ARMv8_4a): New fgroup.
10220 (armv8.4-a): New arch.
10221 * config/arm/arm-tables.opt: Regenerate.
10222 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10223 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10224 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10225 Add matching rules for -march=armv8.4-a and extensions.
10226 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10228 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10230 PR target/81821
10231 * config/rx/rx.md (BW): New mode attribute.
10232 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10234 2018-01-11 Richard Biener <rguenther@suse.de>
10236 PR tree-optimization/83435
10237 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10238 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10239 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10241 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10242 Alan Hayward <alan.hayward@arm.com>
10243 David Sherwood <david.sherwood@arm.com>
10245 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10246 field.
10247 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10248 (aarch64_print_address_internal): Use it to check for a zero offset.
10250 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10251 Alan Hayward <alan.hayward@arm.com>
10252 David Sherwood <david.sherwood@arm.com>
10254 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10255 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10256 Return a poly_int64 rather than a HOST_WIDE_INT.
10257 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10258 rather than a HOST_WIDE_INT.
10259 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10260 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10261 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10262 final_offset from HOST_WIDE_INT to poly_int64.
10263 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10264 to_constant when getting the number of units in an Advanced SIMD
10265 mode.
10266 (aarch64_builtin_vectorized_function): Check for a constant number
10267 of units.
10268 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10269 GET_MODE_SIZE.
10270 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10271 attribute instead of GET_MODE_NUNITS.
10272 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10273 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10274 GET_MODE_SIZE for fixed-size registers.
10275 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10276 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10277 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10278 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10279 (aarch64_print_operand, aarch64_print_address_internal)
10280 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10281 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10282 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10283 Handle polynomial GET_MODE_SIZE.
10284 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10285 wider than SImode without modification.
10286 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10287 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10288 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10289 passing and returning SVE modes.
10290 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10291 rather than GEN_INT.
10292 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10293 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10294 (aarch64_allocate_and_probe_stack_space): Likewise.
10295 (aarch64_layout_frame): Cope with polynomial offsets.
10296 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10297 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10298 polynomial offsets.
10299 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10300 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10301 poly_int64 rather than a HOST_WIDE_INT.
10302 (aarch64_get_separate_components, aarch64_process_components)
10303 (aarch64_expand_prologue, aarch64_expand_epilogue)
10304 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10305 (aarch64_anchor_offset): New function, split out from...
10306 (aarch64_legitimize_address): ...here.
10307 (aarch64_builtin_vectorization_cost): Handle polynomial
10308 TYPE_VECTOR_SUBPARTS.
10309 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10310 GET_MODE_NUNITS.
10311 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10312 number of elements from the PARALLEL rather than the mode.
10313 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10314 rather than GET_MODE_BITSIZE.
10315 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10316 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10317 (aarch64_expand_vec_perm_const_1): Handle polynomial
10318 d->perm.length () and d->perm elements.
10319 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10320 Apply to_constant to d->perm elements.
10321 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10322 polynomial CONST_VECTOR_NUNITS.
10323 (aarch64_move_pointer): Take amount as a poly_int64 rather
10324 than an int.
10325 (aarch64_progress_pointer): Avoid temporary variable.
10326 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10327 the mode attribute instead of GET_MODE.
10329 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10330 Alan Hayward <alan.hayward@arm.com>
10331 David Sherwood <david.sherwood@arm.com>
10333 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10334 x exists before using it.
10335 (aarch64_add_constant_internal): Rename to...
10336 (aarch64_add_offset_1): ...this. Replace regnum with separate
10337 src and dest rtxes. Handle the case in which they're different,
10338 including when the offset is zero. Replace scratchreg with an rtx.
10339 Use 2 additions if there is no spare register into which we can
10340 move a 16-bit constant.
10341 (aarch64_add_constant): Delete.
10342 (aarch64_add_offset): Replace reg with separate src and dest
10343 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10344 Use aarch64_add_offset_1.
10345 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10346 an rtx rather than an int. Take the delta as a poly_int64
10347 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10348 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10349 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10350 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10351 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10352 and aarch64_add_sp.
10353 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10354 aarch64_add_constant.
10356 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10358 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10359 Use scalar_float_mode.
10361 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10363 * config/aarch64/aarch64-simd.md
10364 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10365 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10366 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10367 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10368 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10369 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10370 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10371 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10372 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10373 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10375 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10377 PR target/83514
10378 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10379 targ_options->x_arm_arch_string is non NULL.
10381 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10383 * config/aarch64/aarch64.h
10384 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10386 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10388 PR target/82096
10389 * expmed.c (emit_store_flag_force): Swap if const op0
10390 and change VOIDmode to mode of op0.
10392 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10394 PR rtl-optimization/83761
10395 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10396 than bytes to mode_for_size.
10398 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10400 PR middle-end/83189
10401 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10402 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10403 profile.
10405 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10407 PR middle-end/83575
10408 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10409 when in layout mode.
10410 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10411 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10412 partition fixup.
10414 2018-01-10 Michael Collison <michael.collison@arm.com>
10416 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10417 * config/aarch64/aarch64-option-extension.def: Add
10418 AARCH64_OPT_EXTENSION of 'fp16fml'.
10419 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10420 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10421 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10422 * config/aarch64/constraints.md (Ui7): New constraint.
10423 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10424 (VFMLA_SEL_W): Ditto.
10425 (f16quad): Ditto.
10426 (f16mac1): Ditto.
10427 (VFMLA16_LOW): New int iterator.
10428 (VFMLA16_HIGH): Ditto.
10429 (UNSPEC_FMLAL): New unspec.
10430 (UNSPEC_FMLSL): Ditto.
10431 (UNSPEC_FMLAL2): Ditto.
10432 (UNSPEC_FMLSL2): Ditto.
10433 (f16mac): New code attribute.
10434 * config/aarch64/aarch64-simd-builtins.def
10435 (aarch64_fmlal_lowv2sf): Ditto.
10436 (aarch64_fmlsl_lowv2sf): Ditto.
10437 (aarch64_fmlalq_lowv4sf): Ditto.
10438 (aarch64_fmlslq_lowv4sf): Ditto.
10439 (aarch64_fmlal_highv2sf): Ditto.
10440 (aarch64_fmlsl_highv2sf): Ditto.
10441 (aarch64_fmlalq_highv4sf): Ditto.
10442 (aarch64_fmlslq_highv4sf): Ditto.
10443 (aarch64_fmlal_lane_lowv2sf): Ditto.
10444 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10445 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10446 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10447 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10448 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10449 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10450 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10451 (aarch64_fmlal_lane_highv2sf): Ditto.
10452 (aarch64_fmlsl_lane_highv2sf): Ditto.
10453 (aarch64_fmlal_laneq_highv2sf): Ditto.
10454 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10455 (aarch64_fmlalq_lane_highv4sf): Ditto.
10456 (aarch64_fmlsl_lane_highv4sf): Ditto.
10457 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10458 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10459 * config/aarch64/aarch64-simd.md:
10460 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10461 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10462 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10463 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10464 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10465 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10466 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10467 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10468 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10469 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10470 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10471 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10472 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10473 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10474 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10475 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10476 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10477 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10478 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10479 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10480 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10481 (vfmlsl_low_u32): Ditto.
10482 (vfmlalq_low_u32): Ditto.
10483 (vfmlslq_low_u32): Ditto.
10484 (vfmlal_high_u32): Ditto.
10485 (vfmlsl_high_u32): Ditto.
10486 (vfmlalq_high_u32): Ditto.
10487 (vfmlslq_high_u32): Ditto.
10488 (vfmlal_lane_low_u32): Ditto.
10489 (vfmlsl_lane_low_u32): Ditto.
10490 (vfmlal_laneq_low_u32): Ditto.
10491 (vfmlsl_laneq_low_u32): Ditto.
10492 (vfmlalq_lane_low_u32): Ditto.
10493 (vfmlslq_lane_low_u32): Ditto.
10494 (vfmlalq_laneq_low_u32): Ditto.
10495 (vfmlslq_laneq_low_u32): Ditto.
10496 (vfmlal_lane_high_u32): Ditto.
10497 (vfmlsl_lane_high_u32): Ditto.
10498 (vfmlal_laneq_high_u32): Ditto.
10499 (vfmlsl_laneq_high_u32): Ditto.
10500 (vfmlalq_lane_high_u32): Ditto.
10501 (vfmlslq_lane_high_u32): Ditto.
10502 (vfmlalq_laneq_high_u32): Ditto.
10503 (vfmlslq_laneq_high_u32): Ditto.
10504 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10505 (AARCH64_FL_FOR_ARCH8_4): New.
10506 (AARCH64_ISA_F16FML): New ISA flag.
10507 (TARGET_F16FML): New feature flag for fp16fml.
10508 (doc/invoke.texi): Document new fp16fml option.
10510 2018-01-10 Michael Collison <michael.collison@arm.com>
10512 * config/aarch64/aarch64-builtins.c:
10513 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10514 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10515 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10516 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10517 (AARCH64_ISA_SHA3): New ISA flag.
10518 (TARGET_SHA3): New feature flag for sha3.
10519 * config/aarch64/iterators.md (sha512_op): New int attribute.
10520 (CRYPTO_SHA512): New int iterator.
10521 (UNSPEC_SHA512H): New unspec.
10522 (UNSPEC_SHA512H2): Ditto.
10523 (UNSPEC_SHA512SU0): Ditto.
10524 (UNSPEC_SHA512SU1): Ditto.
10525 * config/aarch64/aarch64-simd-builtins.def
10526 (aarch64_crypto_sha512hqv2di): New builtin.
10527 (aarch64_crypto_sha512h2qv2di): Ditto.
10528 (aarch64_crypto_sha512su0qv2di): Ditto.
10529 (aarch64_crypto_sha512su1qv2di): Ditto.
10530 (aarch64_eor3qv8hi): Ditto.
10531 (aarch64_rax1qv2di): Ditto.
10532 (aarch64_xarqv2di): Ditto.
10533 (aarch64_bcaxqv8hi): Ditto.
10534 * config/aarch64/aarch64-simd.md:
10535 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10536 (aarch64_crypto_sha512su0qv2di): Ditto.
10537 (aarch64_crypto_sha512su1qv2di): Ditto.
10538 (aarch64_eor3qv8hi): Ditto.
10539 (aarch64_rax1qv2di): Ditto.
10540 (aarch64_xarqv2di): Ditto.
10541 (aarch64_bcaxqv8hi): Ditto.
10542 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10543 (vsha512h2q_u64): Ditto.
10544 (vsha512su0q_u64): Ditto.
10545 (vsha512su1q_u64): Ditto.
10546 (veor3q_u16): Ditto.
10547 (vrax1q_u64): Ditto.
10548 (vxarq_u64): Ditto.
10549 (vbcaxq_u16): Ditto.
10550 * config/arm/types.md (crypto_sha512): New type attribute.
10551 (crypto_sha3): Ditto.
10552 (doc/invoke.texi): Document new sha3 option.
10554 2018-01-10 Michael Collison <michael.collison@arm.com>
10556 * config/aarch64/aarch64-builtins.c:
10557 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10558 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10559 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10560 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10561 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10562 (AARCH64_ISA_SM4): New ISA flag.
10563 (TARGET_SM4): New feature flag for sm4.
10564 * config/aarch64/aarch64-simd-builtins.def
10565 (aarch64_sm3ss1qv4si): Ditto.
10566 (aarch64_sm3tt1aq4si): Ditto.
10567 (aarch64_sm3tt1bq4si): Ditto.
10568 (aarch64_sm3tt2aq4si): Ditto.
10569 (aarch64_sm3tt2bq4si): Ditto.
10570 (aarch64_sm3partw1qv4si): Ditto.
10571 (aarch64_sm3partw2qv4si): Ditto.
10572 (aarch64_sm4eqv4si): Ditto.
10573 (aarch64_sm4ekeyqv4si): Ditto.
10574 * config/aarch64/aarch64-simd.md:
10575 (aarch64_sm3ss1qv4si): Ditto.
10576 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10577 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10578 (aarch64_sm4eqv4si): Ditto.
10579 (aarch64_sm4ekeyqv4si): Ditto.
10580 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10581 (sm3part_op): Ditto.
10582 (CRYPTO_SM3TT): Ditto.
10583 (CRYPTO_SM3PART): Ditto.
10584 (UNSPEC_SM3SS1): New unspec.
10585 (UNSPEC_SM3TT1A): Ditto.
10586 (UNSPEC_SM3TT1B): Ditto.
10587 (UNSPEC_SM3TT2A): Ditto.
10588 (UNSPEC_SM3TT2B): Ditto.
10589 (UNSPEC_SM3PARTW1): Ditto.
10590 (UNSPEC_SM3PARTW2): Ditto.
10591 (UNSPEC_SM4E): Ditto.
10592 (UNSPEC_SM4EKEY): Ditto.
10593 * config/aarch64/constraints.md (Ui2): New constraint.
10594 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10595 * config/arm/types.md (crypto_sm3): New type attribute.
10596 (crypto_sm4): Ditto.
10597 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10598 (vsm3tt1aq_u32): Ditto.
10599 (vsm3tt1bq_u32): Ditto.
10600 (vsm3tt2aq_u32): Ditto.
10601 (vsm3tt2bq_u32): Ditto.
10602 (vsm3partw1q_u32): Ditto.
10603 (vsm3partw2q_u32): Ditto.
10604 (vsm4eq_u32): Ditto.
10605 (vsm4ekeyq_u32): Ditto.
10606 (doc/invoke.texi): Document new sm4 option.
10608 2018-01-10 Michael Collison <michael.collison@arm.com>
10610 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10611 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10612 (AARCH64_FL_FOR_ARCH8_4): New.
10613 (AARCH64_FL_V8_4): New flag.
10614 (doc/invoke.texi): Document new armv8.4-a option.
10616 2018-01-10 Michael Collison <michael.collison@arm.com>
10618 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10619 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10620 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10621 * config/aarch64/aarch64-option-extension.def: Add
10622 AARCH64_OPT_EXTENSION of 'sha2'.
10623 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10624 (crypto): Disable sha2 and aes if crypto disabled.
10625 (crypto): Enable aes and sha2 if enabled.
10626 (simd): Disable sha2 and aes if simd disabled.
10627 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10628 New flags.
10629 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10630 (TARGET_SHA2): New feature flag for sha2.
10631 (TARGET_AES): New feature flag for aes.
10632 * config/aarch64/aarch64-simd.md:
10633 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10634 conditional on TARGET_AES.
10635 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10636 (aarch64_crypto_sha1hsi): Make pattern conditional
10637 on TARGET_SHA2.
10638 (aarch64_crypto_sha1hv4si): Ditto.
10639 (aarch64_be_crypto_sha1hv4si): Ditto.
10640 (aarch64_crypto_sha1su1v4si): Ditto.
10641 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10642 (aarch64_crypto_sha1su0v4si): Ditto.
10643 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10644 (aarch64_crypto_sha256su0v4si): Ditto.
10645 (aarch64_crypto_sha256su1v4si): Ditto.
10646 (doc/invoke.texi): Document new aes and sha2 options.
10648 2018-01-10 Martin Sebor <msebor@redhat.com>
10650 PR tree-optimization/83781
10651 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10652 as string arrays.
10654 2018-01-11 Martin Sebor <msebor@gmail.com>
10655 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10657 PR tree-optimization/83501
10658 PR tree-optimization/81703
10660 * tree-ssa-strlen.c (get_string_cst): Rename...
10661 (get_string_len): ...to this. Handle global constants.
10662 (handle_char_store): Adjust.
10664 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10665 Jim Wilson <jimw@sifive.com>
10667 * config/riscv/riscv-protos.h (riscv_output_return): New.
10668 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10669 (riscv_attribute_table, riscv_output_return),
10670 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10671 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10672 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10673 function.
10674 (riscv_expand_prologue): Add early return for naked function.
10675 (riscv_expand_epilogue): Likewise.
10676 (riscv_function_ok_for_sibcall): Return false for naked function.
10677 (riscv_set_current_function): New.
10678 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10679 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10680 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10681 * doc/extend.texi (RISC-V Function Attributes): New.
10683 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10685 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10686 check for 128-bit long double before checking TCmode.
10687 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10688 128-bit long doubles before checking TFmode or TCmode.
10689 (FLOAT128_IBM_P): Likewise.
10691 2018-01-10 Martin Sebor <msebor@redhat.com>
10693 PR tree-optimization/83671
10694 * builtins.c (c_strlen): Unconditionally return zero for the empty
10695 string.
10696 Use -Warray-bounds for warnings.
10697 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10698 for non-constant array indices with COMPONENT_REF, arrays of
10699 arrays, and pointers to arrays.
10700 (gimple_fold_builtin_strlen): Determine and set length range for
10701 non-constant character arrays.
10703 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10705 PR middle-end/81897
10706 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10707 empty blocks.
10709 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10711 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10713 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10715 PR target/83399
10716 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10717 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10718 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10719 indexed_or_indirect_operand predicate.
10720 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10721 (*vsx_le_perm_load_v8hi): Likewise.
10722 (*vsx_le_perm_load_v16qi): Likewise.
10723 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10724 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10725 (*vsx_le_perm_store_v8hi): Likewise.
10726 (*vsx_le_perm_store_v16qi): Likewise.
10727 (eight unnamed splitters): Likewise.
10729 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10731 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10732 * config/rs6000/emmintrin.h: Likewise.
10733 * config/rs6000/mmintrin.h: Likewise.
10734 * config/rs6000/xmmintrin.h: Likewise.
10736 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10738 PR c++/43486
10739 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10740 "public_flag".
10741 * tree.c (tree_nop_conversion): Return true for location wrapper
10742 nodes.
10743 (maybe_wrap_with_location): New function.
10744 (selftest::check_strip_nops): New function.
10745 (selftest::test_location_wrappers): New function.
10746 (selftest::tree_c_tests): Call it.
10747 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10748 (maybe_wrap_with_location): New decl.
10749 (EXPR_LOCATION_WRAPPER_P): New macro.
10750 (location_wrapper_p): New inline function.
10751 (tree_strip_any_location_wrapper): New inline function.
10753 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10755 PR target/83735
10756 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10757 stack_realign_offset for the largest alignment of stack slot
10758 actually used.
10759 (ix86_find_max_used_stack_alignment): New function.
10760 (ix86_finalize_stack_frame_flags): Use it. Set
10761 max_used_stack_alignment if we don't realign stack.
10762 * config/i386/i386.h (machine_function): Add
10763 max_used_stack_alignment.
10765 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10767 * config/arm/arm.opt (-mbranch-cost): New option.
10768 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10769 account.
10771 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10773 PR target/83629
10774 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10775 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10777 2018-01-10 Richard Biener <rguenther@suse.de>
10779 PR debug/83765
10780 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10781 early out so it also covers the case where we have a non-NULL
10782 origin.
10784 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10786 PR tree-optimization/83753
10787 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10788 for non-strided grouped accesses if the number of elements is 1.
10790 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10792 PR target/81616
10793 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10794 * i386.h (TARGET_USE_GATHER): Define.
10795 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10797 2018-01-10 Martin Liska <mliska@suse.cz>
10799 PR bootstrap/82831
10800 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10801 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10802 partitioning.
10803 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10804 CLEANUP_NO_PARTITIONING is not set.
10806 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10808 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10809 for vectors, as a partial revert of r254296.
10810 * rtl.h (const_vec_p): Delete.
10811 (const_vec_duplicate_p): Don't test for vector CONSTs.
10812 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10813 * expmed.c (make_tree): Likewise.
10815 Revert:
10816 * common.md (E, F): Use CONSTANT_P instead of checking for
10817 CONST_VECTOR.
10818 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10819 checking for CONST_VECTOR.
10821 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10823 PR middle-end/83575
10824 * predict.c (force_edge_cold): Handle in more sane way edges
10825 with no prediction.
10827 2018-01-09 Carl Love <cel@us.ibm.com>
10829 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10830 V4SI, V4SF types.
10831 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10832 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10833 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10834 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10835 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10836 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10837 * config/rs6000/rs6000-protos.h: Add extern defition for
10838 rs6000_generate_float2_double_code.
10839 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10840 function.
10841 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10842 (float2_v2df): Add define_expand.
10844 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10846 PR target/83628
10847 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10848 op_mode in the force_to_mode call.
10850 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10852 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10853 instead of checking each element individually.
10854 (aarch64_evpc_uzp): Likewise.
10855 (aarch64_evpc_zip): Likewise.
10856 (aarch64_evpc_ext): Likewise.
10857 (aarch64_evpc_rev): Likewise.
10858 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10859 instead of checking each element individually. Return true without
10860 generating rtl if
10861 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10862 whether all selected elements come from the same input, instead of
10863 checking each element individually. Remove calls to gen_rtx_REG,
10864 start_sequence and end_sequence and instead assert that no rtl is
10865 generated.
10867 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10869 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10870 order of HIGH and CONST checks.
10872 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10874 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10875 if the destination isn't an SSA_NAME.
10877 2018-01-09 Richard Biener <rguenther@suse.de>
10879 PR tree-optimization/83668
10880 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10881 move prologue...
10882 (canonicalize_loop_form): ... here, renamed from ...
10883 (canonicalize_loop_closed_ssa_form): ... this and amended to
10884 swap successor edges for loop exit blocks to make us use
10885 the RPO order we need for initial schedule generation.
10887 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10889 PR tree-optimization/64811
10890 * match.pd: When optimizing comparisons with Inf, avoid
10891 introducing or losing exceptions from comparisons with NaN.
10893 2018-01-09 Martin Liska <mliska@suse.cz>
10895 PR sanitizer/82517
10896 * asan.c (shadow_mem_size): Add gcc_assert.
10898 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10900 Don't save registers in main().
10902 PR target/83738
10903 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10904 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10905 * config/avr/avr.c (avr_set_current_function): Don't error if
10906 naked, OS_task or OS_main are specified at the same time.
10907 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10908 OS_main.
10909 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10910 attribute.
10911 * common/config/avr/avr-common.c (avr_option_optimization_table):
10912 Switch on -mmain-is-OS_task for optimizing compilations.
10914 2018-01-09 Richard Biener <rguenther@suse.de>
10916 PR tree-optimization/83572
10917 * graphite.c: Include cfganal.h.
10918 (graphite_transform_loops): Connect infinite loops to exit
10919 and remove fake edges at the end.
10921 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10923 * ipa-inline.c (edge_badness): Revert accidental checkin.
10925 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10927 PR ipa/80763
10928 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10929 symbols; not inline clones.
10931 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10933 PR target/83507
10934 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10935 hard registers. Formatting fixes.
10937 PR preprocessor/83722
10938 * gcc.c (try_generate_repro): Pass
10939 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10940 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10941 do_report_bug.
10943 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10944 Kito Cheng <kito.cheng@gmail.com>
10946 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10947 (riscv_leaf_function_p): Delete.
10948 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10950 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10952 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10953 function.
10954 (do_ifelse): New function.
10955 (do_isel): New function.
10956 (do_sub3): New function.
10957 (do_add3): New function.
10958 (do_load_mask_compare): New function.
10959 (do_overlap_load_compare): New function.
10960 (expand_compare_loop): New function.
10961 (expand_block_compare): Call expand_compare_loop() when appropriate.
10962 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10963 option description.
10964 (-mblock-compare-inline-loop-limit): New option.
10966 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10968 PR target/83677
10969 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10970 Reverse order of second and third operands in first alternative.
10971 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10972 of first and second elements in UNSPEC_VPERMR vector.
10973 (altivec_expand_vec_perm_le): Likewise.
10975 2018-01-08 Jeff Law <law@redhat.com>
10977 PR rtl-optimizatin/81308
10978 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10979 (process_switch): If group_case_labels makes a change, then set
10980 cfg_altered.
10981 (pass_convert_switch::execute): If a switch is converted, then
10982 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10984 PR rtl-optimization/81308
10985 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10986 splitting insns.
10988 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10990 PR target/83663 - Revert r255946
10991 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10992 generation for cases where splatting a value is not useful.
10993 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10994 across a vec_duplicate and a paradoxical subreg forming a vector
10995 mode to a vec_concat.
10997 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10999 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11000 -march=armv8.3-a variants.
11001 * config/arm/t-multilib: Likewise.
11002 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11004 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11006 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11007 to generate rtl.
11008 (cceq_ior_compare_complement): Give it a name so I can use it, and
11009 change boolean_or_operator predicate to boolean_operator so it can
11010 be used to generate a crand.
11011 (eqne): New code iterator.
11012 (bd/bd_neg): New code_attrs.
11013 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11014 a single define_insn.
11015 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11016 decrement (bdnzt/bdnzf/bdzt/bdzf).
11017 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11018 with the new names of the branch decrement patterns, and added the
11019 names of the branch decrement conditional patterns.
11021 2018-01-08 Richard Biener <rguenther@suse.de>
11023 PR tree-optimization/83563
11024 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11025 cache.
11027 2018-01-08 Richard Biener <rguenther@suse.de>
11029 PR middle-end/83713
11030 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11032 2018-01-08 Richard Biener <rguenther@suse.de>
11034 PR tree-optimization/83685
11035 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11036 references to abnormals.
11038 2018-01-08 Richard Biener <rguenther@suse.de>
11040 PR lto/83719
11041 * dwarf2out.c (output_indirect_strings): Handle empty
11042 skeleton_debug_str_hash.
11043 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11045 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11047 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11048 (emit_store_direct): Likewise.
11049 (arc_trampoline_adjust_address): Likewise.
11050 (arc_asm_trampoline_template): New function.
11051 (arc_initialize_trampoline): Use asm_trampoline_template.
11052 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11053 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11054 * config/arc/arc.md (flush_icache): Delete pattern.
11056 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11058 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11059 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11060 munaligned-access.
11062 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11064 PR target/83681
11065 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11066 by not USED_FOR_TARGET.
11067 (make_pass_resolve_sw_modes): Likewise.
11069 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11071 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11072 USED_FOR_TARGET.
11074 2018-01-08 Richard Biener <rguenther@suse.de>
11076 PR middle-end/83580
11077 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11079 2018-01-08 Richard Biener <rguenther@suse.de>
11081 PR middle-end/83517
11082 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11084 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11086 PR middle-end/81897
11087 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11088 basic blocks with a small number of successors.
11089 (convert_control_dep_chain_into_preds): Improve handling of
11090 forwarder blocks.
11091 (dump_predicates): Split apart into...
11092 (dump_pred_chain): ...here...
11093 (dump_pred_info): ...and here.
11094 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11095 (can_chain_union_be_invalidated_p): Improve check for invalidation
11096 of paths.
11097 (uninit_uses_cannot_happen): Avoid unnecessary if
11098 convert_control_dep_chain_into_preds yielded nothing.
11100 2018-01-06 Martin Sebor <msebor@redhat.com>
11102 PR tree-optimization/83640
11103 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11104 subtracting negative offset from size.
11105 (builtin_access::overlap): Adjust offset bounds of the access to fall
11106 within the size of the object if possible.
11108 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11110 PR rtl-optimization/83699
11111 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11112 extract_bit_field_as_subreg to cases in which the extracted
11113 value is also a vector.
11115 * lra-constraints.c (process_alt_operands): Test for the equivalence
11116 substitutions when detecting a possible reload cycle.
11118 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11120 PR debug/83480
11121 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11122 by default if flag_selective_schedling{,2}. Formatting fixes.
11124 PR rtl-optimization/83682
11125 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11126 if it has non-VECTOR_MODE element mode.
11127 (vec_duplicate_p): Likewise.
11129 PR middle-end/83694
11130 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11131 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11133 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11135 PR target/83604
11136 * config/i386/i386-builtin.def
11137 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11138 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11139 Require also OPTION_MASK_ISA_AVX512F in addition to
11140 OPTION_MASK_ISA_GFNI.
11141 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11142 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11143 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11144 to OPTION_MASK_ISA_GFNI.
11145 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11146 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11147 OPTION_MASK_ISA_AVX512BW.
11148 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11149 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11150 addition to OPTION_MASK_ISA_GFNI.
11151 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11152 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11153 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11154 to OPTION_MASK_ISA_GFNI.
11155 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11156 a requirement for all ISAs rather than any of them with a few
11157 exceptions.
11158 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11159 processing.
11160 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11161 bitmasks to be enabled with 3 exceptions, instead of requiring any
11162 enabled ISA with lots of exceptions.
11163 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11164 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11165 Change avx512bw in isa attribute to avx512f.
11166 * config/i386/sgxintrin.h: Add license boilerplate.
11167 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11168 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11169 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11170 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11171 defined.
11172 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11173 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11174 temporarily sse2 rather than sse if not enabled already.
11176 PR target/83604
11177 * config/i386/sse.md (VI248_VLBW): Rename to ...
11178 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11179 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11180 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11181 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11182 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11183 mode iterator instead of VI248_VLBW.
11185 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11187 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11188 (record_modified): Skip clobbers; add debug output.
11189 (param_change_prob): Use sreal frequencies.
11191 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11193 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11194 punt for user-aligned variables.
11196 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11198 * tree-chrec.c (chrec_contains_symbols): Return true for
11199 POLY_INT_CST.
11201 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11203 PR target/82439
11204 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11205 of (x|y) == x for BICS pattern.
11207 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11209 PR tree-optimization/83605
11210 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11211 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11212 can throw.
11214 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11216 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11217 * config/epiphany/rtems.h: New file.
11219 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11220 Uros Bizjak <ubizjak@gmail.com>
11222 PR target/83554
11223 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11224 QIreg_operand instead of register_operand predicate.
11225 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11226 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11227 comments instead of -fmitigate[-_]rop.
11229 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11231 PR bootstrap/81926
11232 * cgraphunit.c (symbol_table::compile): Switch to text_section
11233 before calling assembly_start debug hook.
11234 * run-rtl-passes.c (run_rtl_passes): Likewise.
11235 Include output.h.
11237 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11239 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11240 range_int_cst_p rather than !symbolic_range_p before calling
11241 extract_range_from_multiplicative_op_1.
11243 2018-01-04 Jeff Law <law@redhat.com>
11245 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11246 redundant test in assertion.
11248 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11250 * doc/rtl.texi: Document machine_mode wrapper classes.
11252 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11254 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11255 using tree_to_uhwi.
11257 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11259 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11260 the VEC_PERM_EXPR fold to fail.
11262 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11264 PR debug/83585
11265 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11266 to switched_sections.
11268 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11270 PR target/83680
11271 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11272 test for d.testing.
11274 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11276 PR target/83387
11277 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11278 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11280 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11282 PR debug/83666
11283 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11284 is BLKmode and bitpos not zero or mode change is needed.
11286 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11288 PR target/83675
11289 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11290 TARGET_VIS2.
11292 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11294 PR target/83628
11295 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11296 instead of MULT rtx. Update all corresponding splitters.
11297 (*saddl_se): Ditto.
11298 (*ssub<modesuffix>): Ditto.
11299 (*ssubl_se): Ditto.
11300 (*cmp_sadd_di): Update split patterns.
11301 (*cmp_sadd_si): Ditto.
11302 (*cmp_sadd_sidi): Ditto.
11303 (*cmp_ssub_di): Ditto.
11304 (*cmp_ssub_si): Ditto.
11305 (*cmp_ssub_sidi): Ditto.
11306 * config/alpha/predicates.md (const23_operand): New predicate.
11307 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11308 Look for ASHIFT, not MULT inner operand.
11309 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11311 2018-01-04 Martin Liska <mliska@suse.cz>
11313 PR gcov-profile/83669
11314 * gcov.c (output_intermediate_file): Add version to intermediate
11315 gcov file.
11316 * doc/gcov.texi: Document new field 'version' in intermediate
11317 file format. Fix location of '-k' option of gcov command.
11319 2018-01-04 Martin Liska <mliska@suse.cz>
11321 PR ipa/82352
11322 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11324 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11326 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11328 2018-01-03 Martin Sebor <msebor@redhat.com>
11330 PR tree-optimization/83655
11331 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11332 checking calls with invalid arguments.
11334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11336 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11337 (vectorizable_mask_load_store): Delete.
11338 (vectorizable_call): Return false for masked loads and stores.
11339 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11340 instead of gimple_assign_rhs1.
11341 (vectorizable_load): Handle IFN_MASK_LOAD.
11342 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11346 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11347 split out from..,
11348 (vectorizable_mask_load_store): ...here.
11349 (vectorizable_load): ...and here.
11351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11353 * tree-vect-stmts.c (vect_build_all_ones_mask)
11354 (vect_build_zero_merge_argument): New functions, split out from...
11355 (vectorizable_load): ...here.
11357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11359 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11360 split out from...
11361 (vectorizable_mask_load_store): ...here.
11362 (vectorizable_store): ...and here.
11364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11366 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11367 split out from...
11368 (vectorizable_mask_load_store): ...here.
11370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11372 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11373 (vect_model_store_cost): Take a vec_load_store_type instead of a
11374 vect_def_type.
11375 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11376 (vect_model_store_cost): Take a vec_load_store_type instead of a
11377 vect_def_type.
11378 (vectorizable_mask_load_store): Update accordingly.
11379 (vectorizable_store): Likewise.
11380 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11384 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11385 IFN_MASK_LOAD calls here rather than...
11386 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11389 Alan Hayward <alan.hayward@arm.com>
11390 David Sherwood <david.sherwood@arm.com>
11392 * expmed.c (extract_bit_field_1): For vector extracts,
11393 fall back to extract_bit_field_as_subreg if vec_extract
11394 isn't available.
11396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11397 Alan Hayward <alan.hayward@arm.com>
11398 David Sherwood <david.sherwood@arm.com>
11400 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11401 they are variable or constant sized.
11402 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11403 slots for constant-sized data.
11405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11406 Alan Hayward <alan.hayward@arm.com>
11407 David Sherwood <david.sherwood@arm.com>
11409 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11410 handling COND_EXPRs with boolean comparisons, try to find a better
11411 basis for the mask type than the boolean itself.
11413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11415 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11416 is calculated and how it can be overridden.
11417 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11418 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11419 if defined.
11420 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11421 if nonzero.
11423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11424 Alan Hayward <alan.hayward@arm.com>
11425 David Sherwood <david.sherwood@arm.com>
11427 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11428 Remove the mode argument.
11429 (aarch64_simd_valid_immediate): Remove the mode and inverse
11430 arguments.
11431 * config/aarch64/iterators.md (bitsize): New iterator.
11432 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11433 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11434 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11435 aarch64_simd_valid_immediate.
11436 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11437 (aarch64_reg_or_bic_imm): Likewise.
11438 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11439 with an insn_type enum and msl with a modifier_type enum.
11440 Replace element_width with a scalar_mode. Change the shift
11441 to unsigned int. Add constructors for scalar_float_mode and
11442 scalar_int_mode elements.
11443 (aarch64_vect_float_const_representable_p): Delete.
11444 (aarch64_can_const_movi_rtx_p)
11445 (aarch64_simd_scalar_immediate_valid_for_move)
11446 (aarch64_simd_make_constant): Update call to
11447 aarch64_simd_valid_immediate.
11448 (aarch64_advsimd_valid_immediate_hs): New function.
11449 (aarch64_advsimd_valid_immediate): Likewise.
11450 (aarch64_simd_valid_immediate): Remove mode and inverse
11451 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11452 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11453 and aarch64_float_const_representable_p on the result.
11454 (aarch64_output_simd_mov_immediate): Remove mode argument.
11455 Update call to aarch64_simd_valid_immediate and use of
11456 simd_immediate_info.
11457 (aarch64_output_scalar_simd_mov_immediate): Update call
11458 accordingly.
11460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11461 Alan Hayward <alan.hayward@arm.com>
11462 David Sherwood <david.sherwood@arm.com>
11464 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11465 (mode_nunits): Likewise CONST_MODE_NUNITS.
11466 * machmode.def (ADJUST_NUNITS): Document.
11467 * genmodes.c (mode_data::need_nunits_adj): New field.
11468 (blank_mode): Update accordingly.
11469 (adj_nunits): New variable.
11470 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11471 parameter.
11472 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11473 listed in adj_nunits.
11474 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11475 listed in adj_nunits. Don't emit case statements for such modes.
11476 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11477 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11478 nothing if adj_nunits is nonnull.
11479 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11480 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11481 (emit_mode_fbit): Update use of print_maybe_const_decl.
11482 (emit_move_size): Likewise. Treat the array as non-const
11483 if adj_nunits.
11484 (emit_mode_adjustments): Handle adj_nunits.
11486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11488 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11489 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11490 (VECTOR_MODES): Use it.
11491 (make_vector_modes): Take the prefix as an argument.
11493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11494 Alan Hayward <alan.hayward@arm.com>
11495 David Sherwood <david.sherwood@arm.com>
11497 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11498 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11499 for MODE_VECTOR_BOOL.
11500 * machmode.def (VECTOR_BOOL_MODE): Document.
11501 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11502 (make_vector_bool_mode): New function.
11503 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11504 MODE_VECTOR_BOOL.
11505 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11506 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11507 Likewise.
11508 * stor-layout.c (int_mode_for_mode): Likewise.
11509 * tree.c (build_vector_type_for_mode): Likewise.
11510 * varasm.c (output_constant_pool_2): Likewise.
11511 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11512 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11513 for MODE_VECTOR_BOOL.
11514 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11515 of mode class checks.
11516 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11517 instead of a list of mode class checks.
11518 (expand_vector_scalar_condition): Likewise.
11519 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11522 Alan Hayward <alan.hayward@arm.com>
11523 David Sherwood <david.sherwood@arm.com>
11525 * machmode.h (mode_size): Change from unsigned short to
11526 poly_uint16_pod.
11527 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11528 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11529 or if measurement_type is not polynomial.
11530 (fixed_size_mode::includes_p): Check for constant-sized modes.
11531 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11532 return a poly_uint16 rather than an unsigned short.
11533 (emit_mode_size): Change the type of mode_size from unsigned short
11534 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11535 (emit_mode_adjustments): Cope with polynomial vector sizes.
11536 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11537 for GET_MODE_SIZE.
11538 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11539 for GET_MODE_SIZE.
11540 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11541 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11542 * caller-save.c (setup_save_areas): Likewise.
11543 (replace_reg_with_saved_mem): Likewise.
11544 * calls.c (emit_library_call_value_1): Likewise.
11545 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11546 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11547 (gen_lowpart_for_combine): Likewise.
11548 * convert.c (convert_to_integer_1): Likewise.
11549 * cse.c (equiv_constant, cse_insn): Likewise.
11550 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11551 (cselib_subst_to_values): Likewise.
11552 * dce.c (word_dce_process_block): Likewise.
11553 * df-problems.c (df_word_lr_mark_ref): Likewise.
11554 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11555 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11556 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11557 (rtl_for_decl_location): Likewise.
11558 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11559 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11560 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11561 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11562 (expand_expr_real_1): Likewise.
11563 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11564 (pad_below): Likewise.
11565 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11566 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11567 * ira.c (get_subreg_tracking_sizes): Likewise.
11568 * ira-build.c (ira_create_allocno_objects): Likewise.
11569 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11570 (ira_sort_regnos_for_alter_reg): Likewise.
11571 * ira-costs.c (record_operand_costs): Likewise.
11572 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11573 (resolve_simple_move): Likewise.
11574 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11575 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11576 (lra_constraints): Likewise.
11577 (CONST_POOL_OK_P): Reject variable-sized modes.
11578 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11579 (add_pseudo_to_slot, lra_spill): Likewise.
11580 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11581 * optabs-query.c (get_best_extraction_insn): Likewise.
11582 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11583 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11584 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11585 * recog.c (offsettable_address_addr_space_p): Likewise.
11586 * regcprop.c (maybe_mode_change): Likewise.
11587 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11588 * regrename.c (build_def_use): Likewise.
11589 * regstat.c (dump_reg_info): Likewise.
11590 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11591 (find_reloads, find_reloads_subreg_address): Likewise.
11592 * reload1.c (eliminate_regs_1): Likewise.
11593 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11594 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11595 (simplify_binary_operation_1, simplify_subreg): Likewise.
11596 * targhooks.c (default_function_arg_padding): Likewise.
11597 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11598 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11599 (verify_gimple_assign_ternary): Likewise.
11600 * tree-inline.c (estimate_move_cost): Likewise.
11601 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11602 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11603 (get_address_cost_ainc): Likewise.
11604 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11605 (vect_supportable_dr_alignment): Likewise.
11606 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11607 (vectorizable_reduction): Likewise.
11608 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11609 (vectorizable_operation, vectorizable_load): Likewise.
11610 * tree.c (build_same_sized_truth_vector_type): Likewise.
11611 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11612 * var-tracking.c (emit_note_insn_var_location): Likewise.
11613 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11614 (ADDR_VEC_ALIGN): Likewise.
11616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11617 Alan Hayward <alan.hayward@arm.com>
11618 David Sherwood <david.sherwood@arm.com>
11620 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11621 unsigned short.
11622 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11623 or if measurement_type is polynomial.
11624 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11625 * combine.c (make_extraction): Likewise.
11626 * dse.c (find_shift_sequence): Likewise.
11627 * dwarf2out.c (mem_loc_descriptor): Likewise.
11628 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11629 (extract_bit_field, extract_low_bits): Likewise.
11630 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11631 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11632 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11633 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11634 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11635 * reload.c (find_reloads): Likewise.
11636 * reload1.c (alter_reg): Likewise.
11637 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11638 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11639 * tree-if-conv.c (predicate_mem_writes): Likewise.
11640 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11641 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11642 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11643 * valtrack.c (dead_debug_insert_temp): Likewise.
11644 * varasm.c (mergeable_constant_section): Likewise.
11645 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11648 Alan Hayward <alan.hayward@arm.com>
11649 David Sherwood <david.sherwood@arm.com>
11651 * expr.c (expand_assignment): Cope with polynomial mode sizes
11652 when assigning to a CONCAT.
11654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11655 Alan Hayward <alan.hayward@arm.com>
11656 David Sherwood <david.sherwood@arm.com>
11658 * machmode.h (mode_precision): Change from unsigned short to
11659 poly_uint16_pod.
11660 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11661 short.
11662 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11663 or if measurement_type is not polynomial.
11664 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11665 in which the mode is already known to be a scalar_int_mode.
11666 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11667 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11668 initializer.
11669 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11670 for GET_MODE_PRECISION.
11671 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11672 for GET_MODE_PRECISION.
11673 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11674 as polynomial.
11675 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11676 (expand_field_assignment, make_extraction): Likewise.
11677 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11678 (get_last_value): Likewise.
11679 * convert.c (convert_to_integer_1): Likewise.
11680 * cse.c (cse_insn): Likewise.
11681 * expr.c (expand_expr_real_1): Likewise.
11682 * lra-constraints.c (simplify_operand_subreg): Likewise.
11683 * optabs-query.c (can_atomic_load_p): Likewise.
11684 * optabs.c (expand_atomic_load): Likewise.
11685 (expand_atomic_store): Likewise.
11686 * ree.c (combine_reaching_defs): Likewise.
11687 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11688 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11689 * tree.h (type_has_mode_precision_p): Likewise.
11690 * ubsan.c (instrument_si_overflow): Likewise.
11692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11693 Alan Hayward <alan.hayward@arm.com>
11694 David Sherwood <david.sherwood@arm.com>
11696 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11697 polynomial numbers of units.
11698 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11699 (valid_vector_subparts_p): New function.
11700 (build_vector_type): Remove temporary shim and take the number
11701 of units as a poly_uint64 rather than an int.
11702 (build_opaque_vector_type): Take the number of units as a
11703 poly_uint64 rather than an int.
11704 * tree.c (build_vector_from_ctor): Handle polynomial
11705 TYPE_VECTOR_SUBPARTS.
11706 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11707 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11708 (build_vector_from_val): If the number of units is variable,
11709 use build_vec_duplicate_cst for constant operands and
11710 VEC_DUPLICATE_EXPR otherwise.
11711 (make_vector_type): Remove temporary is_constant ().
11712 (build_vector_type, build_opaque_vector_type): Take the number of
11713 units as a poly_uint64 rather than an int.
11714 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11715 VECTOR_CST_NELTS.
11716 * cfgexpand.c (expand_debug_expr): Likewise.
11717 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11718 (store_constructor, expand_expr_real_1): Likewise.
11719 (const_scalar_mask_from_tree): Likewise.
11720 * fold-const-call.c (fold_const_reduction): Likewise.
11721 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11722 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11723 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11724 (fold_relational_const): Likewise.
11725 (native_interpret_vector): Likewise. Change the size from an
11726 int to an unsigned int.
11727 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11728 TYPE_VECTOR_SUBPARTS.
11729 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11730 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11731 duplicating a non-constant operand into a variable-length vector.
11732 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11733 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11734 * ipa-icf.c (sem_variable::equals): Likewise.
11735 * match.pd: Likewise.
11736 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11737 * print-tree.c (print_node): Likewise.
11738 * stor-layout.c (layout_type): Likewise.
11739 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11740 * tree-cfg.c (verify_gimple_comparison): Likewise.
11741 (verify_gimple_assign_binary): Likewise.
11742 (verify_gimple_assign_ternary): Likewise.
11743 (verify_gimple_assign_single): Likewise.
11744 * tree-pretty-print.c (dump_generic_node): Likewise.
11745 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11746 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11747 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11748 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11749 (vect_shift_permute_load_chain): Likewise.
11750 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11751 (expand_vector_condition, optimize_vector_constructor): Likewise.
11752 (lower_vec_perm, get_compute_type): Likewise.
11753 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11754 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11755 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11756 (vect_recog_mask_conversion_pattern): Likewise.
11757 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11758 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11759 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11760 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11761 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11762 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11763 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11764 (supportable_widening_operation): Likewise.
11765 (supportable_narrowing_operation): Likewise.
11766 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11767 Likewise.
11768 * varasm.c (output_constant): Likewise.
11770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11771 Alan Hayward <alan.hayward@arm.com>
11772 David Sherwood <david.sherwood@arm.com>
11774 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11775 so that both the length == 3 and length != 3 cases set up their
11776 own permute vectors. Add comments explaining why we know the
11777 number of elements is constant.
11778 (vect_permute_load_chain): Likewise.
11780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11781 Alan Hayward <alan.hayward@arm.com>
11782 David Sherwood <david.sherwood@arm.com>
11784 * machmode.h (mode_nunits): Change from unsigned char to
11785 poly_uint16_pod.
11786 (ONLY_FIXED_SIZE_MODES): New macro.
11787 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11788 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11789 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11790 New typedefs.
11791 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11792 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11793 or if measurement_type is not polynomial.
11794 * genmodes.c (ZERO_COEFFS): New macro.
11795 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11796 poly_uint16.
11797 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11798 Use ZERO_COEFFS when emitting initializers.
11799 * data-streamer.h (bp_pack_poly_value): New function.
11800 (bp_unpack_poly_value): Likewise.
11801 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11802 for GET_MODE_NUNITS.
11803 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11804 for GET_MODE_NUNITS.
11805 * tree.c (make_vector_type): Remove temporary shim and make
11806 the real function take the number of units as a poly_uint64
11807 rather than an int.
11808 (build_vector_type_for_mode): Handle polynomial nunits.
11809 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11810 * emit-rtl.c (const_vec_series_p_1): Likewise.
11811 (gen_rtx_CONST_VECTOR): Likewise.
11812 * fold-const.c (test_vec_duplicate_folding): Likewise.
11813 * genrecog.c (validate_pattern): Likewise.
11814 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11815 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11816 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11817 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11818 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11819 * rtlanal.c (subreg_get_info): Likewise.
11820 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11821 (vect_grouped_load_supported): Likewise.
11822 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11823 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11824 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11825 (simplify_const_unary_operation, simplify_binary_operation_1)
11826 (simplify_const_binary_operation, simplify_ternary_operation)
11827 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11828 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11829 instead of CONST_VECTOR_NUNITS.
11830 * varasm.c (output_constant_pool_2): Likewise.
11831 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11832 explicit-encoded elements in the XVEC for variable-length vectors.
11834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11836 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11839 Alan Hayward <alan.hayward@arm.com>
11840 David Sherwood <david.sherwood@arm.com>
11842 * coretypes.h (fixed_size_mode): Declare.
11843 (fixed_size_mode_pod): New typedef.
11844 * builtins.h (target_builtins::x_apply_args_mode)
11845 (target_builtins::x_apply_result_mode): Change type to
11846 fixed_size_mode_pod.
11847 * builtins.c (apply_args_size, apply_result_size, result_vector)
11848 (expand_builtin_apply_args_1, expand_builtin_apply)
11849 (expand_builtin_return): Update accordingly.
11851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11853 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11854 * cselib.c (cselib_hash_rtx): Likewise.
11855 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11856 CONST_VECTOR encoding.
11858 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11859 Jeff Law <law@redhat.com>
11861 PR target/83641
11862 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11863 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11864 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11865 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11867 PR target/83641
11868 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11869 explicitly probe *sp in a noreturn function if there were any callee
11870 register saves or frame pointer is needed.
11872 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11874 PR debug/83621
11875 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11876 BLKmode for ternary, binary or unary expressions.
11878 PR debug/83645
11879 * var-tracking.c (delete_vta_debug_insn): New inline function.
11880 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11881 insns from get_insns () to NULL instead of each bb separately.
11882 Use delete_vta_debug_insn. No longer static.
11883 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11884 delete_vta_debug_insns callers.
11885 * rtl.h (delete_vta_debug_insns): Declare.
11886 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11887 instead of variable_tracking_main.
11889 2018-01-03 Martin Sebor <msebor@redhat.com>
11891 PR tree-optimization/83603
11892 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11893 arguments past the endof the argument list in functions declared
11894 without a prototype.
11895 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11896 Avoid checking when arguments are null.
11898 2018-01-03 Martin Sebor <msebor@redhat.com>
11900 PR c/83559
11901 * doc/extend.texi (attribute const): Fix a typo.
11902 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11903 issuing -Wsuggest-attribute for void functions.
11905 2018-01-03 Martin Sebor <msebor@redhat.com>
11907 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11908 offset_int::from instead of wide_int::to_shwi.
11909 (maybe_diag_overlap): Remove assertion.
11910 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11911 * gimple-ssa-sprintf.c (format_directive): Same.
11912 (parse_directive): Same.
11913 (sprintf_dom_walker::compute_format_length): Same.
11914 (try_substitute_return_value): Same.
11916 2018-01-03 Jeff Law <law@redhat.com>
11918 PR middle-end/83654
11919 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11920 non-constant residual for zero at runtime and avoid probing in
11921 that case. Reorganize code for trailing problem to mirror handling
11922 of the residual.
11924 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11926 PR tree-optimization/83501
11927 * tree-ssa-strlen.c (get_string_cst): New.
11928 (handle_char_store): Call get_string_cst.
11930 2018-01-03 Martin Liska <mliska@suse.cz>
11932 PR tree-optimization/83593
11933 * tree-ssa-strlen.c: Include tree-cfg.h.
11934 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11935 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11936 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11937 to false.
11938 (strlen_dom_walker::before_dom_children): Call
11939 gimple_purge_dead_eh_edges. Dump tranformation with details
11940 dump flags.
11941 (strlen_dom_walker::before_dom_children): Update call by adding
11942 new argument cleanup_eh.
11943 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11945 2018-01-03 Martin Liska <mliska@suse.cz>
11947 PR ipa/83549
11948 * cif-code.def (VARIADIC_THUNK): New enum value.
11949 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11950 thunks.
11952 2018-01-03 Jan Beulich <jbeulich@suse.com>
11954 * sse.md (mov<mode>_internal): Tighten condition for when to use
11955 vmovdqu<ssescalarsize> for TI and OI modes.
11957 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11959 Update copyright years.
11961 2018-01-03 Martin Liska <mliska@suse.cz>
11963 PR ipa/83594
11964 * ipa-visibility.c (function_and_variable_visibility): Skip
11965 functions with noipa attribure.
11967 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11969 * gcc.c (process_command): Update copyright notice dates.
11970 * gcov-dump.c (print_version): Ditto.
11971 * gcov.c (print_version): Ditto.
11972 * gcov-tool.c (print_version): Ditto.
11973 * gengtype.c (create_file): Ditto.
11974 * doc/cpp.texi: Bump @copying's copyright year.
11975 * doc/cppinternals.texi: Ditto.
11976 * doc/gcc.texi: Ditto.
11977 * doc/gccint.texi: Ditto.
11978 * doc/gcov.texi: Ditto.
11979 * doc/install.texi: Ditto.
11980 * doc/invoke.texi: Ditto.
11982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11984 * vector-builder.h (vector_builder::m_full_nelts): Change from
11985 unsigned int to poly_uint64.
11986 (vector_builder::full_nelts): Update prototype accordingly.
11987 (vector_builder::new_vector): Likewise.
11988 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11989 (vector_builder::operator ==): Likewise.
11990 (vector_builder::finalize): Likewise.
11991 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11992 Take the number of elements as a poly_uint64 rather than an
11993 unsigned int.
11994 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11995 from unsigned int to poly_uint64.
11996 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11997 (vec_perm_indices::new_vector): Likewise.
11998 (vec_perm_indices::length): Likewise.
11999 (vec_perm_indices::nelts_per_input): Likewise.
12000 (vec_perm_indices::input_nelts): Likewise.
12001 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12002 number of elements per input as a poly_uint64 rather than an
12003 unsigned int. Use the original encoding for variable-length
12004 vectors, rather than clamping each individual element.
12005 For the second and subsequent elements in each pattern,
12006 clamp the step and base before clamping their sum.
12007 (vec_perm_indices::series_p): Handle polynomial element counts.
12008 (vec_perm_indices::all_in_range_p): Likewise.
12009 (vec_perm_indices_to_tree): Likewise.
12010 (vec_perm_indices_to_rtx): Likewise.
12011 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12012 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12013 (tree_vector_builder::new_binary_operation): Handle polynomial
12014 element counts. Return false if we need to know the number
12015 of elements at compile time.
12016 * fold-const.c (fold_vec_perm): Punt if the number of elements
12017 isn't known at compile time.
12019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12021 * vec-perm-indices.h (vec_perm_builder): Change element type
12022 from HOST_WIDE_INT to poly_int64.
12023 (vec_perm_indices::element_type): Update accordingly.
12024 (vec_perm_indices::clamp): Handle polynomial element_types.
12025 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12026 (vec_perm_indices::all_in_range_p): Likewise.
12027 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12028 than shwi trees.
12029 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12030 polynomial vec_perm_indices element types.
12031 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12032 * fold-const.c (fold_vec_perm): Likewise.
12033 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12034 * tree-vect-generic.c (lower_vec_perm): Likewise.
12035 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12036 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12037 element type to HOST_WIDE_INT.
12039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12040 Alan Hayward <alan.hayward@arm.com>
12041 David Sherwood <david.sherwood@arm.com>
12043 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12044 rather than an int. Use plus_constant.
12045 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12046 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12049 Alan Hayward <alan.hayward@arm.com>
12050 David Sherwood <david.sherwood@arm.com>
12052 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12053 a HOST_WIDE_INT to a poly_int64.
12055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12056 Alan Hayward <alan.hayward@arm.com>
12057 David Sherwood <david.sherwood@arm.com>
12059 * calls.c (load_register_parameters): Cope with polynomial
12060 mode sizes. Require a constant size for BLKmode parameters
12061 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12062 forces a parameter to be padded at the lsb end in order to
12063 fill a complete number of words, require the parameter size
12064 to be ordered wrt UNITS_PER_WORD.
12066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12067 Alan Hayward <alan.hayward@arm.com>
12068 David Sherwood <david.sherwood@arm.com>
12070 * reload1.c (spill_stack_slot_width): Change element type
12071 from unsigned int to poly_uint64_pod.
12072 (alter_reg): Treat mode sizes as polynomial.
12074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12075 Alan Hayward <alan.hayward@arm.com>
12076 David Sherwood <david.sherwood@arm.com>
12078 * reload.c (complex_word_subreg_p): New function.
12079 (reload_inner_reg_of_subreg, push_reload): Use it.
12081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12082 Alan Hayward <alan.hayward@arm.com>
12083 David Sherwood <david.sherwood@arm.com>
12085 * lra-constraints.c (process_alt_operands): Reject matched
12086 operands whose sizes aren't ordered.
12087 (match_reload): Refer to this check here.
12089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12090 Alan Hayward <alan.hayward@arm.com>
12091 David Sherwood <david.sherwood@arm.com>
12093 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12094 that the mode size is in the set {1, 2, 4, 8, 16}.
12096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12097 Alan Hayward <alan.hayward@arm.com>
12098 David Sherwood <david.sherwood@arm.com>
12100 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12101 Use plus_constant instead of gen_rtx_PLUS.
12103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12104 Alan Hayward <alan.hayward@arm.com>
12105 David Sherwood <david.sherwood@arm.com>
12107 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12108 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12109 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12110 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12111 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12112 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12113 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12114 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12115 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12116 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12117 a poly_int64.
12118 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12119 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12120 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12121 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12122 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12123 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12124 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12125 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12126 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12127 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12128 function.
12129 * expr.c (emit_move_resolve_push): Treat the input and result
12130 of PUSH_ROUNDING as a poly_int64.
12131 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12132 (emit_push_insn): Likewise.
12133 * lra-eliminations.c (mark_not_eliminable): Likewise.
12134 * recog.c (push_operand): Likewise.
12135 * reload1.c (elimination_effects): Likewise.
12136 * rtlanal.c (nonzero_bits1): Likewise.
12137 * calls.c (store_one_arg): Likewise. Require the padding to be
12138 known at compile time.
12140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12141 Alan Hayward <alan.hayward@arm.com>
12142 David Sherwood <david.sherwood@arm.com>
12144 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12145 Use plus_constant instead of gen_rtx_PLUS.
12147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12148 Alan Hayward <alan.hayward@arm.com>
12149 David Sherwood <david.sherwood@arm.com>
12151 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12152 rather than an int.
12154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12155 Alan Hayward <alan.hayward@arm.com>
12156 David Sherwood <david.sherwood@arm.com>
12158 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12159 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12160 via stack temporaries. Treat the mode size as polynomial too.
12162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12163 Alan Hayward <alan.hayward@arm.com>
12164 David Sherwood <david.sherwood@arm.com>
12166 * expr.c (expand_expr_real_2): When handling conversions involving
12167 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12168 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12169 as a poly_uint64 too.
12171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12172 Alan Hayward <alan.hayward@arm.com>
12173 David Sherwood <david.sherwood@arm.com>
12175 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12178 Alan Hayward <alan.hayward@arm.com>
12179 David Sherwood <david.sherwood@arm.com>
12181 * combine.c (can_change_dest_mode): Handle polynomial
12182 REGMODE_NATURAL_SIZE.
12183 * expmed.c (store_bit_field_1): Likewise.
12184 * expr.c (store_constructor): Likewise.
12185 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12186 and polynomial REGMODE_NATURAL_SIZE.
12187 (gen_lowpart_common): Likewise.
12188 * reginfo.c (record_subregs_of_mode): Likewise.
12189 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12196 numbers of elements.
12198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12199 Alan Hayward <alan.hayward@arm.com>
12200 David Sherwood <david.sherwood@arm.com>
12202 * match.pd: Cope with polynomial numbers of vector elements.
12204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12205 Alan Hayward <alan.hayward@arm.com>
12206 David Sherwood <david.sherwood@arm.com>
12208 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12209 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
12216 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12217 (ipa_simd_modify_function_body): Likewise.
12219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12220 Alan Hayward <alan.hayward@arm.com>
12221 David Sherwood <david.sherwood@arm.com>
12223 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12224 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12225 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12226 (expand_vector_condition, vector_element): Likewise.
12227 (subparts_gt): New function.
12228 (get_compute_type): Use subparts_gt.
12229 (count_type_subparts): Delete.
12230 (expand_vector_operations_1): Use subparts_gt instead of
12231 count_type_subparts.
12233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12234 Alan Hayward <alan.hayward@arm.com>
12235 David Sherwood <david.sherwood@arm.com>
12237 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12238 (vect_compile_time_alias): ...this new function. Do the calculation
12239 on poly_ints rather than trees.
12240 (vect_prune_runtime_alias_test_list): Update call accordingly.
12242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12243 Alan Hayward <alan.hayward@arm.com>
12244 David Sherwood <david.sherwood@arm.com>
12246 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12247 numbers of units.
12248 (vect_schedule_slp_instance): Likewise.
12250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12251 Alan Hayward <alan.hayward@arm.com>
12252 David Sherwood <david.sherwood@arm.com>
12254 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12255 constant and extern definitions for variable-length vectors.
12256 (vect_get_constant_vectors): Note that the number of units
12257 is known to be constant.
12259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12260 Alan Hayward <alan.hayward@arm.com>
12261 David Sherwood <david.sherwood@arm.com>
12263 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12264 of units as polynomial. Choose between WIDE and NARROW based
12265 on multiple_p.
12267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12268 Alan Hayward <alan.hayward@arm.com>
12269 David Sherwood <david.sherwood@arm.com>
12271 * tree-vect-stmts.c (simd_clone_subparts): New function.
12272 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12275 Alan Hayward <alan.hayward@arm.com>
12276 David Sherwood <david.sherwood@arm.com>
12278 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12279 vectors as polynomial. Use build_index_vector for
12280 IFN_GOMP_SIMD_LANE.
12282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12283 Alan Hayward <alan.hayward@arm.com>
12284 David Sherwood <david.sherwood@arm.com>
12286 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12287 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12288 for variable-length vectors.
12289 (vectorizable_mask_load_store): Treat the number of units as
12290 polynomial, asserting that it is constant if the condition has
12291 already been enforced.
12292 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
12299 of units as polynomial. Punt if we can't tell at compile time
12300 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
12307 of units as polynomial. Punt on SLP inductions. Use an integer
12308 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12309 cast of such a series for variable-length floating-point
12310 reductions.
12312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12313 Alan Hayward <alan.hayward@arm.com>
12314 David Sherwood <david.sherwood@arm.com>
12316 * tree.h (build_index_vector): Declare.
12317 * tree.c (build_index_vector): New function.
12318 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12319 of units as polynomial, forcibly converting it to a constant if
12320 vectorizable_reduction has already enforced the condition.
12321 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12322 to create a {1,2,3,...} vector.
12323 (vectorizable_reduction): Treat the number of units as polynomial.
12324 Choose vectype_in based on the largest scalar element size rather
12325 than the smallest number of units. Enforce the restrictions
12326 relied on above.
12328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12329 Alan Hayward <alan.hayward@arm.com>
12330 David Sherwood <david.sherwood@arm.com>
12332 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12333 number of units as polynomial.
12335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12336 Alan Hayward <alan.hayward@arm.com>
12337 David Sherwood <david.sherwood@arm.com>
12339 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12340 * target.def (autovectorize_vector_sizes): Return the vector sizes
12341 by pointer, using vector_sizes rather than a bitmask.
12342 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12343 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12344 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12345 Likewise.
12346 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12347 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12348 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12349 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12350 * omp-general.c (omp_max_vf): Likewise.
12351 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12352 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12353 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12354 * tree-vect-slp.c (vect_slp_bb): Likewise.
12355 * doc/tm.texi: Regenerate.
12356 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12357 to a poly_uint64.
12358 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12359 the vector size as a poly_uint64 rather than an unsigned int.
12360 (current_vector_size): Change from an unsigned int to a poly_uint64.
12361 (get_vectype_for_scalar_type): Update accordingly.
12362 * tree.h (build_truth_vector_type): Take the size and number of
12363 units as a poly_uint64 rather than an unsigned int.
12364 (build_vector_type): Add a temporary overload that takes
12365 the number of units as a poly_uint64 rather than an unsigned int.
12366 * tree.c (make_vector_type): Likewise.
12367 (build_truth_vector_type): Take the number of units as a poly_uint64
12368 rather than an unsigned int.
12370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12371 Alan Hayward <alan.hayward@arm.com>
12372 David Sherwood <david.sherwood@arm.com>
12374 * target.def (get_mask_mode): Take the number of units and length
12375 as poly_uint64s rather than unsigned ints.
12376 * targhooks.h (default_get_mask_mode): Update accordingly.
12377 * targhooks.c (default_get_mask_mode): Likewise.
12378 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12379 * doc/tm.texi: Regenerate.
12381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12382 Alan Hayward <alan.hayward@arm.com>
12383 David Sherwood <david.sherwood@arm.com>
12385 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12386 * omp-general.c (omp_max_vf): Likewise.
12387 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12388 (expand_omp_simd): Handle polynomial safelen.
12389 * omp-low.c (omplow_simd_context): Add a default constructor.
12390 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12391 (lower_rec_simd_input_clauses): Update accordingly.
12392 (lower_rec_input_clauses): Likewise.
12394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12395 Alan Hayward <alan.hayward@arm.com>
12396 David Sherwood <david.sherwood@arm.com>
12398 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12399 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12400 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12401 (vect_analyze_slp_cost): Likewise.
12402 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12403 (vect_model_load_cost): Likewise.
12405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12406 Alan Hayward <alan.hayward@arm.com>
12407 David Sherwood <david.sherwood@arm.com>
12409 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12410 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12411 from an unsigned int * to a poly_uint64_pod *.
12412 (calculate_unrolling_factor): New function.
12413 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12416 Alan Hayward <alan.hayward@arm.com>
12417 David Sherwood <david.sherwood@arm.com>
12419 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12420 from an unsigned int to a poly_uint64.
12421 (_loop_vec_info::slp_unrolling_factor): Likewise.
12422 (_loop_vec_info::vectorization_factor): Change from an int
12423 to a poly_uint64.
12424 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12425 (vect_get_num_vectors): New function.
12426 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12427 (vect_get_num_copies): Use vect_get_num_vectors.
12428 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12429 to an unsigned int *.
12430 (vect_analyze_data_refs): Change min_vf from an int * to a
12431 poly_uint64 *.
12432 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12433 than an unsigned HOST_WIDE_INT.
12434 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12435 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12436 to an unsigned int *.
12437 (vect_analyze_data_ref_dependences): Likewise.
12438 (vect_compute_data_ref_alignment): Handle polynomial vf.
12439 (vect_enhance_data_refs_alignment): Likewise.
12440 (vect_prune_runtime_alias_test_list): Likewise.
12441 (vect_shift_permute_load_chain): Likewise.
12442 (vect_supportable_dr_alignment): Likewise.
12443 (dependence_distance_ge_vf): Take the vectorization factor as a
12444 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12445 (vect_analyze_data_refs): Change min_vf from an int * to a
12446 poly_uint64 *.
12447 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12448 vfm1 as a poly_uint64 rather than an int. Make the same change
12449 for the returned bound_scalar.
12450 (vect_gen_vector_loop_niters): Handle polynomial vf.
12451 (vect_do_peeling): Likewise. Update call to
12452 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12453 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12454 be constant.
12455 * tree-vect-loop.c (vect_determine_vectorization_factor)
12456 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12457 (vect_get_known_peeling_cost): Likewise.
12458 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12459 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12460 (vect_transform_loop): Likewise. Use the lowest possible VF when
12461 updating the upper bounds of the loop.
12462 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12463 rather than an int.
12464 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12465 polynomial unroll factors.
12466 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12467 (vect_make_slp_decision): Likewise.
12468 (vect_supported_load_permutation_p): Likewise, and polynomial
12469 vf too.
12470 (vect_analyze_slp_cost): Handle polynomial vf.
12471 (vect_slp_analyze_node_operations): Likewise.
12472 (vect_slp_analyze_bb_1): Likewise.
12473 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12474 than an unsigned HOST_WIDE_INT.
12475 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12476 (vectorizable_load): Handle polynomial vf.
12477 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12478 a poly_uint64.
12479 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12482 Alan Hayward <alan.hayward@arm.com>
12483 David Sherwood <david.sherwood@arm.com>
12485 * match.pd: Handle bit operations involving three constants
12486 and try to fold one pair.
12488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12490 * tree-vect-loop-manip.c: Include gimple-fold.h.
12491 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12492 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12493 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12494 Add a path that uses a step of VF instead of 1, but disable it
12495 for now.
12496 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12497 and niters_no_overflow parameters. Update calls to
12498 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12499 Create a new SSA name if the latter choses to use a ste other
12500 than zero, and return it via niters_vector_mult_vf_var.
12501 * tree-vect-loop.c (vect_transform_loop): Update calls to
12502 vect_do_peeling, vect_gen_vector_loop_niters and
12503 slpeel_make_loop_iterate_ntimes.
12504 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12505 (vect_gen_vector_loop_niters): Update declarations after above changes.
12507 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12509 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12510 128-bit round to integer instructions.
12511 (ceil<mode>2): Likewise.
12512 (btrunc<mode>2): Likewise.
12513 (round<mode>2): Likewise.
12515 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12517 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12518 unaligned VSX load/store on P8/P9.
12519 (expand_block_clear): Allow the use of unaligned VSX
12520 load/store on P8/P9.
12522 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12524 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12525 New function.
12526 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12527 swap associated with both a load and a store.
12529 2018-01-02 Andrew Waterman <andrew@sifive.com>
12531 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12532 * config/riscv/riscv.md (clear_cache): Use it.
12534 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12536 * web.c: Remove out-of-date comment.
12538 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12540 * expr.c (fixup_args_size_notes): Check that any existing
12541 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12542 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12543 (emit_single_push_insn): ...here.
12545 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12547 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12548 (const_vector_encoded_nelts): New function.
12549 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12550 (const_vector_int_elt, const_vector_elt): Declare.
12551 * emit-rtl.c (const_vector_int_elt_1): New function.
12552 (const_vector_elt): Likewise.
12553 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12554 of CONST_VECTOR_ELT.
12556 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12558 * expr.c: Include rtx-vector-builder.h.
12559 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12560 directly on the tree encoding.
12561 (const_vector_from_tree): Likewise.
12562 * optabs.c: Include rtx-vector-builder.h.
12563 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12564 sequence of "u" values.
12565 * vec-perm-indices.c: Include rtx-vector-builder.h.
12566 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12567 directly on the vec_perm_indices encoding.
12569 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12571 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12572 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12573 * rtx-vector-builder.h: New file.
12574 * rtx-vector-builder.c: Likewise.
12575 * rtl.h (rtx_def::u2): Add a const_vector field.
12576 (CONST_VECTOR_NPATTERNS): New macro.
12577 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12578 (CONST_VECTOR_DUPLICATE_P): Likewise.
12579 (CONST_VECTOR_STEPPED_P): Likewise.
12580 (CONST_VECTOR_ENCODED_ELT): Likewise.
12581 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12582 (unwrap_const_vec_duplicate): Likewise.
12583 (const_vec_series_p): Check for a non-duplicated vector encoding.
12584 Say that the function only returns true for integer vectors.
12585 * emit-rtl.c: Include rtx-vector-builder.h.
12586 (gen_const_vec_duplicate_1): Delete.
12587 (gen_const_vector): Call gen_const_vec_duplicate instead of
12588 gen_const_vec_duplicate_1.
12589 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12590 (gen_const_vec_duplicate): Use rtx_vector_builder.
12591 (gen_const_vec_series): Likewise.
12592 (gen_rtx_CONST_VECTOR): Likewise.
12593 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12594 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12595 Build a new vector rather than modifying a CONST_VECTOR in-place.
12596 (handle_special_swappables): Update call accordingly.
12597 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12598 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12599 Build a new vector rather than modifying a CONST_VECTOR in-place.
12600 (handle_special_swappables): Update call accordingly.
12602 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12604 * simplify-rtx.c (simplify_const_binary_operation): Use
12605 CONST_VECTOR_ELT instead of XVECEXP.
12607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12609 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12610 the selector elements to be different from the data elements
12611 if the selector is a VECTOR_CST.
12612 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12613 ssizetype for the selector.
12615 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12617 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12618 before testing each element individually.
12619 * tree-vect-generic.c (lower_vec_perm): Likewise.
12621 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12623 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12624 * selftest-run-tests.c (selftest::run_tests): Call it.
12625 * vector-builder.h (vector_builder::operator ==): New function.
12626 (vector_builder::operator !=): Likewise.
12627 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12628 (vec_perm_indices::all_from_input_p): New function.
12629 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12630 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12631 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12632 instead of reading the VECTOR_CST directly. Detect whether both
12633 vector inputs are the same before constructing the vec_perm_indices,
12634 and update the number of inputs argument accordingly. Use the
12635 utility functions added above. Only construct sel2 if we need to.
12637 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12639 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12640 the broadcast of the low byte.
12641 (expand_mult_highpart): Use an explicit encoding for the permutes.
12642 * optabs-query.c (can_mult_highpart_p): Likewise.
12643 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12644 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12645 (vectorizable_bswap): Likewise.
12646 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12647 explicit encoding for the power-of-2 permutes.
12648 (vect_permute_store_chain): Likewise.
12649 (vect_grouped_load_supported): Likewise.
12650 (vect_permute_load_chain): Likewise.
12652 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12654 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12655 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12656 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12657 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12658 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12659 (vect_gen_perm_mask_any): Likewise.
12661 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12663 * int-vector-builder.h: New file.
12664 * vec-perm-indices.h: Include int-vector-builder.h.
12665 (vec_perm_indices): Redefine as an int_vector_builder.
12666 (auto_vec_perm_indices): Delete.
12667 (vec_perm_builder): Redefine as a stand-alone class.
12668 (vec_perm_indices::vec_perm_indices): New function.
12669 (vec_perm_indices::clamp): Likewise.
12670 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12671 (vec_perm_indices::new_vector): New function.
12672 (vec_perm_indices::new_expanded_vector): Update for new
12673 vec_perm_indices class.
12674 (vec_perm_indices::rotate_inputs): New function.
12675 (vec_perm_indices::all_in_range_p): Operate directly on the
12676 encoded form, without computing elided elements.
12677 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12678 encoding. Update for new vec_perm_indices class.
12679 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12680 the given vec_perm_builder.
12681 (expand_vec_perm_var): Update vec_perm_builder constructor.
12682 (expand_mult_highpart): Use vec_perm_builder instead of
12683 auto_vec_perm_indices.
12684 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12685 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12686 or double series encoding as appropriate.
12687 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12688 vec_perm_indices instead of auto_vec_perm_indices.
12689 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12690 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12691 (vect_permute_store_chain): Likewise.
12692 (vect_grouped_load_supported): Likewise.
12693 (vect_permute_load_chain): Likewise.
12694 (vect_shift_permute_load_chain): Likewise.
12695 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12696 (vect_transform_slp_perm_load): Likewise.
12697 (vect_schedule_slp_instance): Likewise.
12698 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12699 (vectorizable_mask_load_store): Likewise.
12700 (vectorizable_bswap): Likewise.
12701 (vectorizable_store): Likewise.
12702 (vectorizable_load): Likewise.
12703 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12704 vec_perm_indices instead of auto_vec_perm_indices. Use
12705 tree_to_vec_perm_builder to read the vector from a tree.
12706 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12707 vec_perm_builder instead of a vec_perm_indices.
12708 (have_whole_vector_shift): Use vec_perm_builder and
12709 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12710 truncation to calc_vec_perm_mask_for_shift.
12711 (vect_create_epilog_for_reduction): Likewise.
12712 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12713 from auto_vec_perm_indices to vec_perm_indices.
12714 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12715 instead of changing individual elements.
12716 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12717 the vector in d.perm.
12718 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12719 from auto_vec_perm_indices to vec_perm_indices.
12720 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12721 instead of changing individual elements.
12722 (arm_vectorize_vec_perm_const): Use new_vector to install
12723 the vector in d.perm.
12724 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12725 Update vec_perm_builder constructor.
12726 (rs6000_expand_interleave): Likewise.
12727 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12728 (rs6000_expand_interleave): Likewise.
12730 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12732 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12733 to qimode could truncate the indices.
12734 * optabs.c (expand_vec_perm_var): Likewise.
12736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12738 * Makefile.in (OBJS): Add vec-perm-indices.o.
12739 * vec-perm-indices.h: New file.
12740 * vec-perm-indices.c: Likewise.
12741 * target.h (vec_perm_indices): Replace with a forward class
12742 declaration.
12743 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12744 * optabs.h: Include vec-perm-indices.h.
12745 (expand_vec_perm): Delete.
12746 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12747 (expand_vec_perm_const): Declare.
12748 * target.def (vec_perm_const_ok): Replace with...
12749 (vec_perm_const): ...this new hook.
12750 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12751 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12752 * doc/tm.texi: Regenerate.
12753 * optabs.def (vec_perm_const): Delete.
12754 * doc/md.texi (vec_perm_const): Likewise.
12755 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12756 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12757 expand_vec_perm for constant permutation vectors. Assert that
12758 the mode of variable permutation vectors is the integer equivalent
12759 of the mode that is being permuted.
12760 * optabs-query.h (selector_fits_mode_p): Declare.
12761 * optabs-query.c: Include vec-perm-indices.h.
12762 (selector_fits_mode_p): New function.
12763 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12764 is defined, instead of checking whether the vec_perm_const_optab
12765 exists. Use targetm.vectorize.vec_perm_const instead of
12766 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12767 fit in the vector mode before using a variable permute.
12768 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12769 vec_perm_indices instead of an rtx.
12770 (expand_vec_perm): Replace with...
12771 (expand_vec_perm_const): ...this new function. Take the selector
12772 as a vec_perm_indices rather than an rtx. Also take the mode of
12773 the selector. Update call to shift_amt_for_vec_perm_mask.
12774 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12775 Use vec_perm_indices::new_expanded_vector to expand the original
12776 selector into bytes. Check whether the indices fit in the vector
12777 mode before using a variable permute.
12778 (expand_vec_perm_var): Make global.
12779 (expand_mult_highpart): Use expand_vec_perm_const.
12780 * fold-const.c: Includes vec-perm-indices.h.
12781 * tree-ssa-forwprop.c: Likewise.
12782 * tree-vect-data-refs.c: Likewise.
12783 * tree-vect-generic.c: Likewise.
12784 * tree-vect-loop.c: Likewise.
12785 * tree-vect-slp.c: Likewise.
12786 * tree-vect-stmts.c: Likewise.
12787 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12788 Delete.
12789 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12790 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12791 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12792 (aarch64_vectorize_vec_perm_const): ...this new function.
12793 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12794 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12795 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12796 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12797 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12798 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12799 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12800 into...
12801 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12802 check for NEON modes.
12803 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12804 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12805 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12806 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12807 into...
12808 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12809 the old VEC_PERM_CONST conditions.
12810 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12811 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12812 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12813 (ia64_vectorize_vec_perm_const_ok): Merge into...
12814 (ia64_vectorize_vec_perm_const): ...this new function.
12815 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12816 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12817 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12818 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12819 * config/mips/mips.c (mips_expand_vec_perm_const)
12820 (mips_vectorize_vec_perm_const_ok): Merge into...
12821 (mips_vectorize_vec_perm_const): ...this new function.
12822 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12823 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12824 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12825 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12826 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12827 (rs6000_expand_vec_perm_const): Delete.
12828 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12829 Delete.
12830 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12831 (altivec_expand_vec_perm_const_le): Take each operand individually.
12832 Operate on constant selectors rather than rtxes.
12833 (altivec_expand_vec_perm_const): Likewise. Update call to
12834 altivec_expand_vec_perm_const_le.
12835 (rs6000_expand_vec_perm_const): Delete.
12836 (rs6000_vectorize_vec_perm_const_ok): Delete.
12837 (rs6000_vectorize_vec_perm_const): New function.
12838 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12839 an element count and rtx array.
12840 (rs6000_expand_extract_even): Update call accordingly.
12841 (rs6000_expand_interleave): Likewise.
12842 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12843 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12844 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12845 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12846 (rs6000_expand_vec_perm_const): Delete.
12847 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12848 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12849 (altivec_expand_vec_perm_const_le): Take each operand individually.
12850 Operate on constant selectors rather than rtxes.
12851 (altivec_expand_vec_perm_const): Likewise. Update call to
12852 altivec_expand_vec_perm_const_le.
12853 (rs6000_expand_vec_perm_const): Delete.
12854 (rs6000_vectorize_vec_perm_const_ok): Delete.
12855 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12856 reference to the SPE evmerge intructions.
12857 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12858 an element count and rtx array.
12859 (rs6000_expand_extract_even): Update call accordingly.
12860 (rs6000_expand_interleave): Likewise.
12861 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12862 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12863 new function.
12864 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12866 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12868 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12869 vector mode and that that mode matches the mode of the data
12870 being permuted.
12871 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12872 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12873 directly using expand_vec_perm_1 when forcing selectors into
12874 registers.
12875 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12877 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12879 * optabs-query.h (can_vec_perm_p): Delete.
12880 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12881 * optabs-query.c (can_vec_perm_p): Split into...
12882 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12883 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12884 particular selector is valid.
12885 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12886 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12887 (vect_grouped_load_supported): Likewise.
12888 (vect_shift_permute_load_chain): Likewise.
12889 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12890 (vect_transform_slp_perm_load): Likewise.
12891 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12892 (vectorizable_bswap): Likewise.
12893 (vect_gen_perm_mask_checked): Likewise.
12894 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12895 implementations of variable permutation vectors into account
12896 when deciding which selector to use.
12897 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12898 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12899 with a false third argument.
12900 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12901 to test whether the constant selector is valid and can_vec_perm_var_p
12902 to test whether a variable selector is valid.
12904 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12906 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12907 * optabs-query.c (can_vec_perm_p): Likewise.
12908 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12909 instead of vec_perm_indices.
12910 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12911 (vect_gen_perm_mask_checked): Likewise,
12912 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12913 (vect_gen_perm_mask_checked): Likewise,
12915 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12917 * optabs-query.h (qimode_for_vec_perm): Declare.
12918 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12919 (qimode_for_vec_perm): ...this new function.
12920 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12922 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12924 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12925 does not have a conditional at the top.
12927 2018-01-02 Richard Biener <rguenther@suse.de>
12929 * ipa-inline.c (big_speedup_p): Fix expression.
12931 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12933 PR target/81616
12934 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12935 for generic 4->6.
12937 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12939 PR target/81616
12940 Generic tuning.
12941 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12942 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12943 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12944 cond_taken_branch_cost 3->4.
12946 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12948 PR tree-optimization/83581
12949 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12950 TODO_cleanup_cfg if any changes have been made.
12952 PR middle-end/83608
12953 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12954 convert_modes if target mode has the right side, but different mode
12955 class.
12957 PR middle-end/83609
12958 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12959 last argument when extracting from CONCAT. If either from_real or
12960 from_imag is NULL, use expansion through memory. If result is not
12961 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12962 the parts directly to inner mode, if even that fails, use expansion
12963 through memory.
12965 PR middle-end/83623
12966 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12967 check for bswap in mode rather than HImode and use that in expand_unop
12968 too.
12969 \f
12970 Copyright (C) 2018 Free Software Foundation, Inc.
12972 Copying and distribution of this file, with or without modification,
12973 are permitted in any medium without royalty provided the copyright
12974 notice and this notice are preserved.