1 2017-04-03 Eric Botcazou <ebotcazou@adacore.com>
3 * tree-nested.c (get_descriptor_type): Make sure that the alignment of
4 descriptors is at least equal to that of functions.
6 2017-04-02 Uros Bizjak <ubizjak@gmail.com>
8 * config/i386/sse.md (movdi_to_sse): Add missing DONE.
10 2017-04-02 Uros Bizjak <ubizjak@gmail.com>
13 * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
14 (mov<IMOD4:mode>): New expander.
15 (*mov<IMOD4:mode>_internal): New insn and split pattern.
17 2017-03-31 Segher Boessenkool <segher@kernel.crashing.org>
19 PR rtl-optimization/79405
20 * fwprop.c (propagations_left): New variable.
21 (forward_propagate_into): Decrement it.
22 (fwprop_init): Initialize it.
23 (fw_prop): If the variable has reached zero, stop propagating.
26 2017-03-31 Jakub Jelinek <jakub@redhat.com>
29 * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
30 a FUNCTION_DECL, pass it as decl instead of origin to
33 2017-03-31 Alexander Monakov <amonakov@ispras.ru>
35 * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
38 2017-03-31 Pat Haugen <pthaugen@us.ibm.com>
41 * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
42 TARGET_VSX_SMALL_INTEGER.
44 2017-03-31 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
46 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
47 reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
49 2017-03-31 Matthew Fortune <matthew.fortune@imgtec.com>
51 * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
52 extraction from odd-numbered MSA register.
54 2017-03-31 Jakub Jelinek <jakub@redhat.com>
57 * expmed.c (store_bit_field_1): Don't attempt to create
58 a word subreg out of hard registers wider than word if they
59 have HARD_REGNO_NREGS of 1 for their mode.
62 * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
63 conversions to integer types wider than word and pointer.
66 * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
67 (rtx_equal_for_cselib_p): Pass 0 to it.
68 * cselib.c (cselib_hasher::equal): Likewise.
69 (rtx_equal_for_cselib_1): Add depth argument. If depth
70 is 128, don't look up VALUE locs and punt. Increment
71 depth in recursive calls when walking VALUE locs.
73 2017-03-31 Bernd Edlinger <bernd.edlinger@hotmail.de>
75 * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
76 (make_gcov_file_name): Use the canonical path name for generating
78 (read_line): Fix handling of files with ascii null bytes.
80 2017-03-30 Matthew Fortune <matthew.fortune@imgtec.com>
82 * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
83 to initialise a vector register instead
86 2017-03-30 Jakub Jelinek <jakub@redhat.com>
89 * gimplify.c (omp_default_clause): Use %qs instead of %s in
92 2017-03-30 Peter Bergner <bergner@vnet.ibm.com>
95 * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
96 (dfp_diex_<mode>): Update mode of operand 1.
97 * doc/extend.texi (dxex, dxexq): Document change to return type.
98 (diex, diexq): Document change to argument type.
100 2017-03-30 Martin Jambor <mjambor@suse.cz>
103 * cgraph.h (cgraph_build_function_type_skip_args): Declare.
104 * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
105 it reflects the signature changes performed at the callee side.
106 * cgraphclones.c (build_function_type_skip_args): Make public, renamed
107 to cgraph_build_function_type_skip_args.
108 (build_function_decl_skip_args): Adjust call to the above function.
110 2017-03-30 Jakub Jelinek <jakub@redhat.com>
114 (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
115 register as dest whenever it is a MEM not rtx_equal_p to the
116 corresponding dup operand, and when forcing into reg move the
117 reg into the memory afterwards.
118 (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
119 Likewise. Use <ssehalfvecmode> instead of <ssequartermode>
120 for the force_reg mode.
121 (avx512vl_vextractf128<mode>): Use register as dest either
122 always when a MEM, or when it is a MEM not rtx_equal_p to the
123 corresponding dup operand, or even not when it is a CONST_VECTOR
124 depending on the mode and lo vs. hi.
125 (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
127 (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
128 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
129 Likewise. Require that operands[2] is even.
130 (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
131 Remove extraneous parens. Require that operands[2] is a multiple
133 (vec_extract_lo_<mode><mask_name>): Don't bother testing if
134 operands[0] is a MEM if <mask_applied>, the predicates/constraints
135 disallow memory then.
137 2017-03-30 Richard Biener <rguenther@suse.de>
139 PR tree-optimization/77498
140 * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
141 to non-constants over backedges.
143 2017-03-29 Segher Boessenkool <segher@kernel.crashing.org>
145 PR rtl-optimization/80233
146 * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
147 as last_combined_insn. Do not test for BARRIER_P separately.
149 2017-03-29 Andreas Schwab <schwab@suse.de>
152 * calls.c (prepare_call_address): Convert funexp to Pmode before
155 2017-03-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
157 PR tree-optimization/80158
158 * gimple-ssa-strength-reduction.c (replace_mult_candidate):
159 Handle possible future case of more than one alternate
161 (replace_rhs_if_not_dup): Likewise.
162 (replace_one_candidate): Likewise.
164 2017-03-28 Vladimir Makarov <vmakarov@redhat.com>
166 PR rtl-optimization/80193
167 * ira.c (ira): Do not check allocation for LRA.
169 2017-03-28 Alexander Monakov <amonakov@ispras.ru>
171 * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
172 (nvptx_output_simt_exit): Declare.
173 * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
174 cfun->machine->unisimt_location. Handle NULL unisimt_predicate.
175 (init_softstack_frame): Move initialization of crtl->is_leaf to...
176 (nvptx_declare_function_name): ...here. Emit declaration of local
177 memory space buffer for omp_simt_enter insn.
178 (nvptx_output_unisimt_switch): New.
179 (nvptx_output_softstack_switch): New.
180 (nvptx_output_simt_enter): New.
181 (nvptx_output_simt_exit): New.
182 * config/nvptx/nvptx.h (struct machine_function): New fields
183 has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
184 * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
185 (UNSPECV_SIMT_EXIT): Ditto.
186 (omp_simt_enter_insn): New insn.
187 (omp_simt_enter): New expansion.
188 (omp_simt_exit): New insn.
189 * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
191 * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
192 (expand_GOMP_SIMT_ENTER_ALLOC): New.
193 (expand_GOMP_SIMT_EXIT): New.
194 * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
195 (GOMP_SIMT_ENTER_ALLOC): Ditto.
196 (GOMP_SIMT_EXIT): Ditto.
197 * target-insns.def (omp_simt_enter): New insn.
198 (omp_simt_exit): Ditto.
199 * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
201 (lower_rec_simd_input_clauses): Implement SIMT privatization.
202 (lower_rec_input_clauses): Likewise.
203 (lower_lastprivate_clauses): Handle SIMT privatization.
205 * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
206 (ompdevlow_adjust_simt_enter): New.
207 (find_simtpriv_var_op): New.
208 (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
209 IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
211 * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
212 * tree-inline.c (expand_call_inline): Handle SIMT privatization.
213 (copy_decl_for_dup_finish): Ditto.
215 * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
217 2017-03-28 Uros Bizjak <ubizjak@gmail.com>
220 * config/i386/i386.c (ix86_option_override_internal): Always
221 allow -mpreferred-stack-boundary=3 for 64-bit targets.
223 2017-03-28 Bin Cheng <bin.cheng@arm.com>
225 * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
227 2017-03-28 Bin Cheng <bin.cheng@arm.com>
229 * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
230 mark new edge's irreducible flag accordign to it.
231 (vect_do_peeling): Check loop preheader edge's irreducible flag
232 and pass it to function slpeel_add_loop_guard.
234 2017-03-28 Richard Sandiford <richard.sandiford@arm.com>
236 PR tree-optimization/80218
237 * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
238 Update block frequencies and counts.
240 2017-03-28 Richard Biener <rguenther@suse.de>
242 PR tree-optimization/78644
243 * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
244 of a simplification result we may not use it at all.
246 2017-03-28 Richard Biener <rguenther@suse.de>
249 * tree-inline.c (copy_phis_for_bb): Do not create PHI node
250 without arguments, generate default definition of a SSA name.
252 2017-03-28 Richard Biener <rguenther@suse.de>
255 * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
256 TYPE_REF_CAN_ALIAS_ALL references.
257 * fold-const.c (fold_indirect_ref_1): Likewise.
259 2017-03-28 Martin Liska <mliska@suse.cz>
262 * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
263 thunk call as DECL_GIMPLE_REG_P when vector or complex type.
265 2017-03-28 Claudiu Zissulescu <claziss@synopsys.com>
266 Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
268 * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
269 (EXTRA_SPECS): Define.
270 (SUBTARGET_EXTRA_SPECS): Likewise.
271 (SUBTARGET_CPP_SPEC): Likewise.
272 * config/arc/elf.h (EXTRA_SPECS): Renamed to
273 SUBTARGET_EXTRA_SPECS.
274 * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
276 2017-03-28 Claudiu Zissulescu <claziss@synopsys.com>
278 * config/arc/simdext.md (vst64_insn): Update pattern.
279 (vld32wh_insn): Likewise.
280 (vld32wl_insn): Likewise.
281 (vld64_insn): Likewise.
282 (vld32_insn): Likewise.
284 2017-03-28 Marek Polacek <polacek@redhat.com>
287 * fold-const.c (fold_comparison): Use protected_set_expr_location
288 instead of SET_EXPR_LOCATION.
290 2017-03-28 Markus Trippelsdorf <markus@trippelsdorf.de>
292 * tree.c (add_expr): Avoid name lookup warning.
294 2017-03-27 Jeff Law <law@redhat.com>
296 PR tree-optimization/80216
297 * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
298 function name. Limit recursion depth.
299 (record_temporary_equivalences): Corresponding changes.
301 2017-03-27 Jonathan Wakely <jwakely@redhat.com>
303 * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
306 2017-03-27 Jakub Jelinek <jakub@redhat.com>
309 * reg-notes.def (REG_CFA_NOTE): Define. Use it for CFA related
311 * cfgcleanup.c (reg_note_cfa_p): New array.
312 (insns_have_identical_cfa_notes): New function.
313 (old_insns_match_p): Don't cross-jump in between /f
314 and non-/f instructions. If both i1 and i2 are frame related,
315 verify all CFA notes, their order and content.
317 2017-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
320 * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
321 HImode and SImode with zero extend to DImode to one insn.
322 (bswap<mode>2_extenddi): Likewise.
323 (bswapsi2_extenddi): Likewise.
324 (bswaphi2_extendsi): Likewise.
325 (bswaphi2): Combine bswap HImode and SImode into one insn.
326 Separate memory insns from swapping register.
327 (bswapsi2): Likewise.
328 (bswap<mode>2): Likewise.
329 (bswaphi2_internal): Delete, no longer used.
330 (bswapsi2_internal): Likewise.
331 (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
332 store, and gpr<-gpr swap insns.
333 (bswap<mode>2_store): Likewise.
334 (bswaphi2_reg): Register only splitter, combine with the splitter.
335 (bswaphi2 splitter): Likewise.
336 (bswapsi2_reg): Likewise.
337 (bswapsi2 splitter): Likewise.
338 (bswapdi2): If we have the LDBRX and STDBRX instructions, split
339 the insns into load, store, and register/register insns.
340 (bswapdi2_ldbrx): Likewise.
341 (bswapdi2_load): Likewise.
342 (bswapdi2_store): Likewise.
343 (bswapdi2_reg): Likewise.
345 2017-03-27 Gunther Nikl <gnikl@users.sourceforge.net>
347 * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
348 (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
350 2017-03-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
353 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
355 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
356 special handling for target option conflicts between dform
357 options (-mpower9-dform, -mpower9-dform-vector,
358 -mpower9-dform-scalar) and -mno-direct-move.
360 2017-03-27 Richard Biener <rguenther@suse.de>
362 PR tree-optimization/80181
363 * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
365 2017-03-27 Claudiu Zissulescu <claziss@synopsys.com>
367 * config/arc/predicates.md (move_double_src_operand): Replace the
368 call to move_double_src_operand with a call to address_operand.
370 2017-03-27 Claudiu Zissulescu <claziss@synopsys.com>
372 * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
373 * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
374 * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
376 2017-03-27 Claudiu Zissulescu <claziss@synopsys.com>
378 * config/arc/predicates.md (long_immediate_loadstore_operand):
379 Consider scaled addresses cases.
381 2017-03-27 Claudiu Zissulescu <claziss@synopsys.com>
383 * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
384 restored when in interrupt.
385 * config/arc/arc.md (simple_return): ARCv2 rtie instruction
386 doesn't have delay slot.
388 2017-03-27 Richard Biener <rguenther@suse.de>
391 * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
392 inlined thunk clones.
394 2017-03-27 Jakub Jelinek <jakub@redhat.com>
397 * asan.c (instrument_derefs): Copy over last operand from
398 original COMPONENT_REF to the new COMPONENT_REF with
399 DECL_BIT_FIELD_REPRESENTATIVE.
400 * ubsan.c (instrument_object_size): Likewise.
402 2017-03-27 Richard Biener <rguenther@suse.de>
404 PR tree-optimization/80170
405 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
406 sure DR/SCEV didnt fold in constants we do not see when looking
407 at the reference base alignment.
409 2017-03-27 Richard Biener <rguenther@suse.de>
412 * gimple-fold.c (fold_ctor_reference): Properly guard against
413 NULL return value from canonicalize_constructor_val.
415 2017-03-25 Uros Bizjak <ubizjak@gmail.com>
418 * config/i386/i386.c (ix86_expand_builtin)
419 <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
420 flags reg setting and flags reg using instructions.
421 <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto. Use non-flags reg
422 clobbering instructions to zero extend op2.
424 2017-03-25 Gerald Pfeifer <gerald@pfeifer.com>
426 * doc/install.texi (Configuration) <--with-aix-soname>:
427 Update link to AIX ld.
429 2017-03-25 Bernd Schmidt <bschmidt@redhat.com>
431 PR rtl-optimization/80160
432 PR rtl-optimization/80159
433 * lra-assigns.c (must_not_spill_p): Tighten new test to also take
434 reg_alternate_class into account.
436 2017-03-24 Vladimir Makarov <vmakarov@redhat.com>
439 * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
440 to consider in curr_insn_transform.
442 2017-03-24 Jakub Jelinek <jakub@redhat.com>
444 * genrecog.c (validate_pattern): Add VEC_SELECT validation.
445 * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
448 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
450 * config/s390/s390-builtins.def: Add VXE builtins. Add a flags
451 argument to the overloaded builtin variants. Use the new flag to
452 deprecate certain builtin variants.
453 * config/s390/s390-builtin-types.def: Add new builtin types.
454 * config/s390/s390-builtins.h: Support new flags field for
456 * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
457 (s390_macro_to_expand): Enable vector float data type.
458 (s390_cpu_cpp_builtins_internal): Indicate support of the new
459 builtins by incrementing the __VEC__ version number.
460 (s390_expand_overloaded_builtin): Support expansion of vec_xl and
462 (s390_resolve_overloaded_builtin): Emit error messages depending
463 on the builtin flags.
464 * config/s390/s390.c (s390_expand_builtin): Support additional
465 flags argument. Change error message to match the messages
467 * config/s390/s390.md: New UNSPEC_* constants.
468 (op_type): Add new instruction types.
469 * config/s390/vecintrin.h: Add new builtins and test data class
471 * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
472 (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
473 (VEC_INEXACT, VEC_NOINEXACT): New constants.
474 ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
475 ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
476 ("vec_mergel<mode>"): V_HW -> VEC_HW.
478 ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
479 ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
480 ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
481 ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
483 ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
484 ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
485 ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
486 ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
488 ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
489 ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
490 ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
491 ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
492 ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
493 ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
494 ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
496 ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
497 ("vec_scatter_element<V_HW_4:mode>_DI")
498 ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
499 ("vec_fpint<mode>", "vflls")
500 ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
501 ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
502 ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
503 ("*vec_cmphe<mode>_cc"): ... these.
505 ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
506 mode constant instead of magic value.
508 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
510 * config/s390/s390.c (s390_expand_vec_compare): Support other
511 vector floating point modes than just V2DF.
512 (s390_expand_vcond): Likewise.
513 (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
514 (s390_cannot_change_mode_class): Prevent mode changes between TF
515 and V1TF in vector registers.
516 * config/s390/s390.md (DF, SF): New mode attributes.
517 ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
518 ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
519 SFmode support for VRs.
520 * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
522 (VFT, VF_HW): New mode iterators.
523 (vw, sdx): New mode attributes.
524 ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
525 ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
526 ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
527 ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
528 ("vec_unorderedv2df"): Adjust the v2df only patterns to support
529 also the new vector floating point modes. Renaming to ...
531 ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
532 ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
533 ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
534 ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
535 ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
536 ("vec_unordered<mode>"): ... these.
538 ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
539 ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
540 ("*vec_extendv2df"): New insn definitions.
542 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
544 * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
545 ("mulditi3_2", "*muldi3_sign"): New patterns.
546 ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
547 rename the pattern definition.
549 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
551 * config/s390/s390.md ("indirect_jump"): Turn insn definition into
553 ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
555 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
557 * config/s390/s390.c (s390_expand_vec_init): Use vllezl
558 instruction if possible.
559 * config/s390/vector.md (vec_halfnumelts): New mode
561 ("*vec_vllezlf<mode>"): New pattern.
563 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
565 * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
566 ("popcountv4si2", "popcountv2di2"): Rename to ...
567 ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
568 ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
570 ("popcount<mode>2_vxe"): New pattern.
572 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
574 * common/config/s390/s390-common.c (processor_flags_table): Add
576 * config.gcc: Add arch12.
577 * config/s390/driver-native.c (s390_host_detect_local_cpu):
578 Default to arch12 for unknown CPU model numbers.
579 * config/s390/s390-builtins.def: Add B_VXE builtin flag.
580 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
581 PROCESSOR_max sanity check.
582 * config/s390/s390-opts.h (enum processor_type): Add
584 * config/s390/s390.c (processor_table): Add arch12.
585 (s390_expand_builtin): Add check for B_VXE flag.
586 (s390_issue_rate): Add PROCESSOR_ARCH12.
587 (s390_get_sched_attrmask): Likewise.
588 (s390_get_unit_mask): Likewise.
589 (s390_sched_score): Enable z13 scheduling for arch12.
590 (s390_sched_reorder): Likewise.
591 (s390_sched_variable_issue): Likewise.
592 * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
594 (s390_tune_attr): Use z13 scheduling also for arch12.
595 (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
596 (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
597 (TARGET_VXE_P): New macros.
598 * config/s390/s390.md: Add arch12 to cpu attribute. Add arch12
599 and vxe to cpu_facility. Add arch12 and vxe to enabled attribute.
600 * config/s390/s390.opt: Add arch12 as processor_type.
602 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
604 * config/s390/s390.md
605 ("fixuns_truncdddi2", "fixuns_trunctddi2")
606 ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
607 ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
609 ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
610 Rename expanders to ...
612 ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
613 ("fixuns_truncdddi2_emu"): ... these.
615 ("fixuns_trunc<mode>si2_emu"): New expander.
617 ("*fixuns_truncdfdi2_z13"): Rename to ...
618 ("*fixuns_truncdfdi2_vx"): ... this.
620 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
622 * config/s390/2964.md: Remove the single element vector compare
623 instructions which are no longer used.
624 * config/s390/s390.c (s390_select_ccmode): Remove handling of
626 (s390_canonicalize_comparison): Remove handling of DFmode
628 (s390_expand_vec_compare_scalar): Remove function.
629 (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
630 * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
632 ("*cmp<mode>_ccs"): Add wfcdb instruction.
634 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
636 * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
638 ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
639 will anyway by matched by mov<mode>_64dfp.
641 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
643 * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
644 vlef/vstef. Add missing operand to vleif.
646 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
648 * config/s390/s390.c (s390_expand_vec_init): Enable vector load
649 pair for all vector types with 64 bit elements.
650 * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
651 * config/s390/vector.md (V_HW_64): ... here.
652 (V_128_NOSINGLE): New mode iterator.
653 ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
654 ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
655 ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
656 ("*vec_load_pairv2di"): Change to ...
657 ("*vec_load_pair<mode>"): ... this one.
659 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
661 * config/s390/constraints.md: Add comments.
662 (jKK): Reject element sizes > 8 bytes.
663 * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
665 * config/s390/s390.md: Add the s_operand checks formerly in
666 s390_split_ok_p to various splitters where they are still
668 * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
669 for 128 bit vectors. Plus two splitters.
671 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
673 * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
676 2017-03-24 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
679 * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
680 error if the boundary argument is not constant.
682 2017-03-24 Jakub Jelinek <jakub@redhat.com>
684 PR rtl-optimization/80112
685 * loop-doloop.c (doloop_condition_get): Don't check condition
686 if cmp isn't SET with IF_THEN_ELSE src.
688 2017-03-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
690 PR tree-optimization/80158
691 * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
692 replacing a candidate statement, also replace it for the
693 candidate's alternate interpretation.
694 (replace_rhs_if_not_dup): Likewise.
695 (replace_one_candidate): Likewise.
697 2017-03-24 Richard Biener <rguenther@suse.de>
699 PR tree-optimization/80167
700 * graphite-isl-ast-to-gimple.c
701 (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
703 (translate_isl_ast_to_gimple::get_rename): Likewise.
705 2017-03-23 Kelvin Nilsen <kelvin@gcc.gnu.org>
707 * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
708 handling of certain combinations of target options, including the
709 combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
710 -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
712 2017-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
715 * config/arm/arm.md (*load_multiple): Add reload_completed to
718 2017-03-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
719 Richard Biener <rguenth@suse.de>
721 PR tree-optimization/79908
722 PR tree-optimization/80136
723 * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
724 been cast away, gimplify_and_add suffices.
726 2017-03-23 Markus Trippelsdorf <markus@trippelsdorf.de>
728 * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
730 2017-03-23 Richard Biener <rguenther@suse.de>
732 PR tree-optimization/80032
733 * gimplify.c (gimple_push_cleanup): Forced unconditional
734 cleanups still have to go to the conditional_cleanups
737 2017-03-22 Jakub Jelinek <jakub@redhat.com>
739 PR tree-optimization/80072
740 * tree-ssa-reassoc.c (struct operand_entry): Change id field type
742 (next_operand_entry_id): Change type to unsigned int.
743 (sort_by_operand_rank): Make sure to return the right return value
744 even if unsigned fields are bigger than INT_MAX.
745 (struct oecount): Change cnt and id type to unsigned int.
746 (oecount_hasher::equal): Formatting fix.
747 (oecount_cmp): Make sure to return the right return value
748 even if unsigned fields are bigger than INT_MAX.
749 (undistribute_ops_list): Change next_oecount_id type to unsigned int.
752 * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
753 TREE_READONLY on result if writing it more than once.
756 * doc/invoke.texi (-fsanitize=thread): Document that with
757 -fnon-call-exceptions atomics are not able to throw
761 * tsan.c: Include tree-eh.h.
762 (instrument_builtin_call): Call maybe_clean_eh_stmt or
763 maybe_clean_or_replace_eh_stmt where needed.
764 (instrument_memory_accesses): Add cfg_changed argument.
765 Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
767 (tsan_pass): Adjust caller. Return TODO_cleanup_cfg if cfg_changed.
769 PR rtl-optimization/63191
770 * config/i386/i386.c (ix86_delegitimize_address): Turn into small
771 wrapper function, moved the whole old content into ...
772 (ix86_delegitimize_address_1): ... this. New inline function.
773 (ix86_find_base_term): Use ix86_delegitimize_address_1 with
774 true as last argument instead of ix86_delegitimize_address.
776 2017-03-22 Wilco Dijkstra <wdijkstr@arm.com>
778 * config/aarch64/aarch64.c (generic_branch_cost): Copy
779 cortexa57_branch_cost.
781 2017-03-22 Wilco Dijkstra <wdijkstr@arm.com>
783 * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
785 2017-03-21 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
788 * doc/md.texi (Constraints): Document wA constraint.
789 * config/rs6000/constraints.md (wA): New.
790 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
791 (rs6000_init_hard_regno_mode_ok): Init wA constraint.
792 * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
793 * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
795 2017-03-22 Cesar Philippidis <cesar@codesourcery.com>
798 * gimplify.c (is_oacc_declared): New function.
799 (oacc_default_clause): Use it to set default flags for acc declared
800 variables inside parallel regions.
801 (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
803 (gimplify_oacc_declare): Gimplify the declare clauses. Add the
804 declare attribute to any decl as necessary.
806 2017-03-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
809 * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
810 (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
811 * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
812 (arm_arch_lpae): This.
813 * config/arm/arm.c (arm_arch7ve): Rename into ...
814 (arm_arch_lpae): This. Define it in term of isa_bit_lpae.
815 * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
818 2017-03-22 Martin Liska <mliska@suse.cz>
821 * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
822 error message instead of an ICE.
824 2017-03-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
826 * doc/extend.texi (6.11 Additional Floating Types): Revise.
828 2017-03-21 Kelvin Nilsen <kelvin@gcc.gnu.org>
830 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
832 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
835 2017-03-21 Martin Sebor <msebor@redhat.com>
837 * doc/extend.texi: Use "cannot" instead of "can't."
838 * doc/hostconfig.texi: Same.
839 * doc/install.texi: Same.
840 * doc/invoke.texi: Same.
841 * doc/loop.texi: Same.
843 * doc/objc.texi: Same.
844 * doc/rtl.texi: Same.
846 * doc/tm.texi.in: Same.
847 * doc/trouble.texi: Same.
849 2017-03-21 Alexandre Oliva <aoliva@redhat.com>
852 * dwarf2out.c (struct checksum_attributes): Add at_alignment.
853 (collect_checksum_attributes): Set it.
854 (die_checksum_ordered): Use it.
856 2017-03-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
858 PR tree-optimization/79908
859 * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
860 change: For a VA_ARG whose LHS has been cast away, use
861 force_gimple_operand to construct the side effects.
863 2017-03-21 David Malcolm <dmalcolm@redhat.com>
866 * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
867 more amenable to translation.
868 (oacc_loop_auto_partitions): Likewise.
870 2017-03-21 Marek Polacek <polacek@redhat.com>
871 Martin Sebor <msebor@redhat.com>
873 PR tree-optimization/80109
874 * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
877 2017-03-21 Jakub Jelinek <jakub@redhat.com>
878 Segher Boessenkool <segher@kernel.crashing.org>
881 * combine.c (can_combine_p): Revert the 2017-03-20 change, only
882 check reg_used_between_p between insn and one of succ or succ2
883 depending on if succ is artificial insn not inserted into insn
886 2017-03-21 Martin Liska <mliska@suse.cz>
888 PR gcov-profile/80081
889 * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
890 * doc/gcc.texi: Include gcov-dump stuff.
891 * doc/gcov-dump.texi: New file.
893 2017-03-21 Toma Tabacu <toma.tabacu@imgtec.com>
895 PR rtl-optimization/79150
896 * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
897 conditional jump, if the jump is the last insn of the loop.
899 2017-03-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
900 Richard Biener <rguenth@suse.de>
902 PR tree-optimization/79908
903 * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
904 been cast away, use force_gimple_operand to construct the side
907 2017-03-21 Martin Liska <mliska@suse.cz>
910 * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
913 2017-03-21 Brad Spengler <spender@grsecurity.net>
916 * plugin.c (htab_hash_plugin): New function.
917 (add_new_plugin): Use it and adjust.
918 (parse_plugin_arg_opt): Adjust.
919 (init_one_plugin): Likewise.
921 2017-03-21 Richard Biener <rguenther@suse.de>
923 PR tree-optimization/80032
924 * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
925 if set force the cleanup to happen unconditionally.
926 (gimplify_target_expr): Push inserted clobbers with force_uncond
927 to avoid them being removed by control-dependent DCE.
929 2017-03-21 Richard Biener <rguenther@suse.de>
931 PR tree-optimization/80122
932 * tree-inline.c (copy_bb): Do not expans va-arg packs or
933 va_arg_pack_len when the inlined call stmt requires pack
935 * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
937 2017-03-21 Jakub Jelinek <jakub@redhat.com>
940 * tsan.c (instrument_builtin_call): If the memory model argument
941 is not a constant, assume it is valid.
944 * fold-const.c (round_up_loc): Negate divisor in unsigned type to
947 2017-03-20 Segher Boessenkool <segher@kernel.crashing.org>
949 PR rtl-optimization/79910
950 * combine.c (can_combine_p): Do not allow combining an I0 or I1
951 if its dest is used by an insn before I2 (other than the combined
952 insns themselves, which are properly handled already).
954 2017-03-20 Segher Boessenkool <segher@kernel.crashing.org>
957 2017-03-17 Bernd Schmidt <bschmidt@redhat.com>
959 * combine.c (record_used_regs): New static function.
960 (try_combine): Handle situations where there is an additional
961 instruction between I2 and I3 which needs to have a LOG_LINK
965 2017-03-17 Jim Wilson <jim.wilson@linaro.org>
967 * combine.c (try_combine): Delete redundant i1 test. Call
968 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
970 2017-03-20 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
973 * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
976 2017-03-20 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
978 PR tree-optimization/80054
979 * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
980 the optimization if a PHI or any of its arguments is not dominated
981 by the candidate's basis. Use gphi* rather than gimple* as
983 (replace_profitable_candidates): Clean up a gimple* variable that
984 should be a gphi* variable.
986 2017-03-20 Martin Sebor <msebor@redhat.com>
989 * doc/extend.texi (attribute constructor): Document present limitation.
991 2017-03-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
994 * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
995 __POWER9_VECTOR__ #ifdef control, change template definition to
996 use Power9-specific built-in function.
997 (vec_any_eq): Likewise.
998 * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
999 to control outcomes from this test.
1000 (vector_ae_<mode>p): For VEC_F modes, likewise.
1002 2017-03-20 Ian Lance Taylor <iant@google.com>
1004 * config/i386/i386.c (ix86_function_regparm): Save an extra
1005 register for -fsplit-stack with DECL_STATIC_CHAIN.
1007 2017-03-17 Palmer Dabbelt <palmer@dabbelt.com>
1010 * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
1011 (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
1013 2017-03-17 Palmer Dabbelt <palmer@dabbelt.com>
1015 * config/riscv/riscv.c (riscv_print_operand): Use "fence
1017 * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
1020 2017-03-20 Marek Polacek <polacek@redhat.com>
1023 * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
1025 2017-03-20 Richard Biener <rguenther@suse.de>
1027 PR tree-optimization/80113
1028 * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
1029 allocate extra SSA name for PHI def.
1030 (add_close_phis_to_outer_loops): Likewise.
1031 (add_close_phis_to_merge_points): Likewise.
1032 (copy_loop_close_phi_args): Likewise.
1033 (copy_cond_phi_nodes): Likewise.
1035 2017-03-20 Martin Liska <mliska@suse.cz>
1038 * tree-chkp.c (chkp_build_returned_bound): Do not build
1039 returned bounds for a LHS that's not a BOUNDED_P type.
1041 2017-03-20 Martin Liska <mliska@suse.cz>
1045 * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
1046 COMPLEX_CST and VECTOR_CST.
1048 2017-03-20 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1051 * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
1052 target operand. A new splitter adds the clobber statement in case
1053 the target operand is dead anyway.
1055 2017-03-19 Gerald Pfeifer <gerald@pfeifer.com>
1057 * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
1058 to age-old versions of binutils and glibc.
1060 2017-03-18 Segher Boessenkool <segher@kernel.crashing.org>
1062 * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
1064 2017-03-18 Gerald Pfeifer <gerald@pfeifer.com>
1066 * doc/contrib.texi (Contributors): Add Segher Boessenkool.
1068 2017-03-18 Gerald Pfeifer <gerald@pfeifer.com>
1070 * doc/install.texi (Specific) <arm-*-eabi>: Remove old
1071 requirement for binutils 2.13.
1073 2017-03-17 Jim Wilson <jim.wilson@linaro.org>
1075 * combine.c (try_combine): Delete redundant i1 test. Call
1076 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
1078 2017-03-17 Palmer Dabbelt <palmer@dabbelt.com
1080 * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
1081 riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
1083 <riscv64-*-elf>: Re-arrange section
1084 <riscv32-*-elf>: Add a note about requiring binutils 2.28.
1085 <riscv32-*-linux>: Likewise.
1086 <riscv64-*-elf>: Likewise
1087 <riscv64-*-linux>: Likewise.
1089 2017-03-17 Richard Earnshaw <rearnsha@arm.com>
1092 * aarch64.opt(verbose-cost-dump): Fix typo.
1094 2017-03-17 Pat Haugen <pthaugen@us.ibm.com>
1097 * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
1098 for VECTOR_UNIT_VSX_P (<MODE>mode) too.
1100 2017-03-17 Bernd Schmidt <bschmidt@redhat.com>
1102 * reload.c (find_reloads): When reloading a nonoffsettable address,
1103 use RELOAD_OTHER for it and its address reloads.
1105 PR rtl-optimization/79910
1106 * combine.c (record_used_regs): New static function.
1107 (try_combine): Handle situations where there is an additional
1108 instruction between I2 and I3 which needs to have a LOG_LINK
1111 2017-03-17 Jeff Law <law@redhat.com>
1113 PR tree-optimization/71437
1114 * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
1115 conditional in the hash table first.
1116 (vrp_dom_walker::before_dom_children): Extract condition from
1117 ASSERT_EXPR. Record condition, its inverion and any implied
1120 2017-03-17 Marek Polacek <polacek@redhat.com>
1121 Markus Trippelsdorf <markus@trippelsdorf.de>
1123 PR tree-optimization/80079
1124 * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
1127 2017-03-17 Richard Biener <rguenther@suse.de>
1130 * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
1131 Properly verify the LHS before the RHS possibly claims to be
1133 (stmt_could_throw_p): Hande gimple conds fully here. Clobbers
1136 2017-03-17 Martin Jambor <mjambor@suse.cz>
1138 * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
1139 (List of -O2 options): Likewise.
1140 (-fipa-bit-cp): Replace "ipa" with "interprocedural."
1143 2017-03-17 Tom de Vries <tom@codesourcery.com>
1145 * gcov-dump.c (print_usage): Print bug_report_url.
1147 2017-03-17 Richard Biener <rguenther@suse.de>
1150 * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
1151 (parser::peek): Likewise.
1153 2017-03-17 Richard Biener <rguenther@suse.de>
1155 PR tree-optimization/80048
1156 * sese.c (free_sese_info): Properly release rename_map and
1157 copied_bb_map elements.
1159 2017-03-16 Alexandre Oliva <aoliva@redhat.com>
1161 * gimple-ssa-store-merging.c (struct imm_store_chain_info):
1162 Add linked-list forward and backlinks. Insert on
1163 construction, remove on destruction.
1164 (class pass_store_merging): Add m_stores_head field.
1165 (pass_store_merging::terminate_and_process_all_chains):
1166 Iterate over m_stores_head list.
1167 (pass_store_merging::terminate_all_aliasing_chains):
1169 (pass_store_merging::execute): Check for debug stmts first.
1170 Push new chains onto the m_stores_head stack.
1172 2017-03-16 Michael Meissner <meissner@linux.vnet.ibm.com>
1175 * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
1176 SPLAT operation on ISA 2.07 64-bit systems that have direct move,
1177 but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
1179 2017-03-16 Jeff Law <law@redhat.com>
1181 PR tree-optimization/71437
1182 * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
1183 member function. Implementation moved into after_dom_children
1184 member function and into the threader's thread_outgoing_edges
1186 (dom_opt_dom_walker::after_dom_children): Simplify by moving
1187 some code into new thread_outgoing_edges.
1188 * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
1189 definition. Simplify marker handling (do it here). Assume we always
1190 have the available expression and the const/copies tables.
1191 (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
1193 * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
1194 * tree-vrp.c (equiv_stack): No longer file scoped.
1195 (vrp_dom_walker): New class.
1196 (vrp_dom_walker::before_dom_children): New member function.
1197 (vrp_dom_walker::after_dom_children): Likewise.
1198 (identify_jump_threads): Setup domwalker. Use it rather than
1199 walking edges in a random order by hand. Simplify setup/finalization.
1200 (finalize_jump_threads): Remove.
1201 (vrp_finalize): Do not call identify_jump_threads here.
1202 (execute_vrp): Do it here instead and call thread_through_all_blocks
1205 PR tree-optimization/71437
1206 * tree-ssa-dom.c (pfn_simplify): Add basic_block argument. All
1208 (simplify_stmt_for_jump_threading): Add basic_block argument. All
1210 (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
1211 (dom_opt_dom_walker::thread_across_edge): Remove
1212 handle_dominating_asserts argument. All callers changed.
1213 (record_temporary_equivalences_from_stmts_at_dest): Corresponding
1214 changes. Remove calls to lhs_of_dominating_assert. Other
1215 uses of handle_dominating_asserts turn into unconditional code
1216 (simplify_control_stmt_condition_1): Likewise.
1217 (simplify_control_stmt_condition): Likewise.
1218 (thread_through_normal_block, thread_across_edge): Likewise.
1219 * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
1220 * tree-vrp.c (lhs_of_dominating_assert): Move here. Return original
1221 object if it is not an SSA_NAME.
1222 (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
1223 before calling into the VRP specific simplifiers.
1224 (identify_jump_threads): Remove handle_dominating_asserts
1227 2017-03-16 Jakub Jelinek <jakub@redhat.com>
1230 * tree-diagnostic.c (default_tree_printer): No longer static.
1231 * tree-diagnostic.h (default_tree_printer): New prototype.
1233 2017-03-16 Tamar Christina <tamar.christina@arm.com>
1235 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
1236 Change ins into fmov.
1238 2017-03-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1240 * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
1241 * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
1242 Use h_con constraint for operand 1.
1243 (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
1244 (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
1246 2017-03-15 Jeff Law <law@redhat.com>
1248 PR tree-optimization/71437
1249 * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
1250 (record_temporary_equivalences): Use it.
1252 PR tree-optimization/71437
1253 * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
1254 tree-ssa-scopedtables.
1255 (lookup_avail_expr, build_and_record_new_cond): Likewise.
1256 (record_conditions, record_cond, vuse_eq): Likewise.
1257 (record_edge_info): Adjust to API tweak of record_conditions.
1258 (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
1259 (record_temporary_equivalences, optimize_stmt): Likewise.
1260 (eliminate_redundant_computations): Likewise.
1261 (record_equivalences_from_stmt): Likewise.
1262 * tree-ssa-scopedtables.c: Include options.h and params.h.
1263 (vuse_eq): New function, moved from tree-ssa-dom.c
1264 (build_and_record_new_cond): Likewise.
1265 (record_conditions): Likewise. Accept vector of conditions rather
1266 than edge_equivalence structure for first argument.
1267 for the first argument.
1268 (avail_exprs_stack::lookup_avail_expr): New member function, moved
1269 from tree-ssa-dom.c.
1270 (avail_exprs_stack::record_cond): Likewise.
1271 * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
1272 from tree-ssa-dom.c.
1273 (avail_exprs_stack): Add new member functions lookup_avail_expr
1275 (record_conditions): Declare.
1277 2017-03-15 Vladimir Makarov <vmakarov@redhat.com>
1280 * lra-constraints.c (process_alt_operands): Increase reject for
1281 reloading an input/output operand.
1283 2017-03-15 Michael Meissner <meissner@linux.vnet.ibm.com>
1286 * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
1287 insns to convert from signed/unsigned char/short to IEEE 128-bit
1289 (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
1291 2017-03-15 Uros Bizjak <ubizjak@gmail.com>
1294 * config/i386/i386.c (ix86_vector_duplicate_value): Create
1295 subreg of inner mode for values already in registers.
1297 2017-03-15 Bernd Schmidt <bschmidt@redhat.com>
1299 * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
1300 iteration reg is used after the loop.
1302 2017-03-14 Martin Sebor <msebor@redhat.com>
1304 PR tree-optimization/79800
1305 * gimple-ssa-sprintf.c (format_floating: Add argument. Handle
1306 precision in negative-positive range.
1307 (format_floating): Call non-const overload with adjusted precision.
1309 2017-03-14 Michael Meissner <meissner@linux.vnet.ibm.com>
1312 * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
1315 2017-03-14 Martin Sebor <msebor@redhat.com>
1318 * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
1319 * builtins.def (aligned_alloc): Use it.
1322 * Makefile.in (GTFILES): Add calls.c.
1323 * calls.c: Include "gt-calls.h".
1325 2017-03-14 Bernd Schmidt <bschmidt@redhat.com>
1327 PR rtl-optimization/79728
1328 * regs.h (struct target_regs): New field
1329 x_contains_allocatable_regs_of_mode.
1330 (contains_allocatable_regs_of_mode): New macro.
1331 * reginfo.c (init_reg_sets_1): Initialize it, and change
1332 contains_reg_of_mode so it includes global regs as well.
1333 * reload.c (push_reload): Use contains_allocatable_regs_of_mode
1334 rather than contains_regs_of_mode.
1336 2017-03-14 Martin Liska <mliska@suse.cz>
1338 * doc/invoke.texi: Document options that can't be combined with
1339 -fcheck-pointer-bounds.
1341 2017-03-14 Martin Liska <mliska@suse.cz>
1344 * doc/invoke.texi (-Wchkp): Document the option.
1346 2017-03-14 Martin Liska <mliska@suse.cz>
1348 * Makefile.in: Install gcov-dump.
1350 2017-03-14 Martin Liska <mliska@suse.cz>
1352 * multiple_target.c (expand_target_clones): Bail out for
1353 an invalid attribute.
1355 2017-03-14 Richard Biener <rguenther@suse.de>
1357 * alias.c (struct alias_set_entry): Pack properly.
1358 * cfgloop.h (struct loop): Likewise.
1359 * cse.c (struct set): Likewise.
1360 * ipa-utils.c (struct searchc_env): Likewise.
1361 * loop-invariant.c (struct invariant): Likewise.
1362 * lra-remat.c (struct cand): Likewise.
1363 * recog.c (struct change_t): Likewise.
1364 * rtl.h (struct address_info): Likewise.
1365 * symbol-summary.h (function_summary): Likewise.
1366 * tree-loop-distribution.c (struct partition): Likewise.
1367 * tree-object-size.c (struct object_size_info): Likewise.
1368 * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
1369 * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
1370 * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
1371 * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
1372 * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
1373 * tree-vectorizer.h (struct _loop_vec_info): Likewise.
1374 (struct _stmt_vec_info): Likewise.
1376 2017-03-14 Martin Liska <mliska@suse.cz>
1379 * multiple_target.c (create_dispatcher_calls): Check that
1380 a target can create a function dispatcher.
1382 2017-03-14 Martin Liska <mliska@suse.cz>
1385 * multiple_target.c (expand_target_clones): Drop local.local
1386 flag for default implementation.
1388 2017-03-14 Richard Biener <rguenther@suse.de>
1390 PR tree-optimization/80030
1391 * tree-vect-stmts.c (vectorizable_store): Plug memleak.
1393 2017-03-13 Kito Cheng <kito.cheng@gmail.com>
1395 * config/riscv/riscv.c (riscv_emit_float_compare>: Use
1396 gcc_fallthrough() instead of __attribute__((fallthrough));
1398 2017-03-13 Gerald Pfeifer <gerald@pfeifer.com>
1400 * doc/gcc.texi: Remove "up" link to (DIR).
1401 * doc/gccint.texi: Ditto.
1403 2017-03-13 Gerald Pfeifer <gerald@pfeifer.com>
1405 * doc/install.texi (Specific) <avr>: Remove reference to
1408 2017-03-13 Jeff Law <law@redhat.com>
1410 * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
1411 attribute rather than comments.
1413 * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
1414 match_scratch operand is highest.
1416 2017-03-13 Martin Liska <mliska@suse.cz>
1419 * ipa-pure-const.c (warn_function_noreturn): If the declarations
1420 is a CHKP clone, use original declaration.
1422 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
1424 * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
1425 (arc_conditional_register_usage): Use a different allocation order
1426 when optimizing for size.
1427 * common/config/arc/arc-common.c (arc_option_optimization_table):
1428 Section anchors default on when optimizing for size.
1430 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
1432 * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
1434 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
1436 * config/arc/arc.c (arc_output_addsi): Emit code density adds.
1437 * config/arc/arc.md (cpu_facility): Add cd variant.
1438 (*movqi_insn): Add code density variant.
1439 (*movhi_insn): Likewise.
1440 (*movqi_insn): Likewise.
1441 (*addsi3_mixed): Likewise.
1442 (subsi3_insn): Likewise.
1444 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
1446 * config/arc/arc.md (movsi_cond_exec): Update constraint.
1448 2017-03-13 Claudiu Zissulescu <claziss@synopsys.com>
1450 * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
1451 expressions with MINUS and UNARY ops.
1453 2017-03-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1456 * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
1458 (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
1459 between vec_select and vector argument.
1460 (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
1461 (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
1462 (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
1463 (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
1464 (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
1465 (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
1467 2017-03-13 Richard Biener <rguenther@suse.de>
1470 * params.def (vect-max-peeling-for-alignment): Fix typo.
1472 2017-03-12 Gerald Pfeifer <gerald@pfeifer.com>
1474 * doc/install.texi (Specific) <mips-*-*>: Remove description of
1475 issue that only occurred with binutils below 2.18.
1477 2017-03-12 Gerald Pfeifer <gerald@pfeifer.com>
1479 * doc/install.texi (Specific) <cris-axis-elf>: No longer
1480 refer to binutils 2.11/2.12 minimum.
1482 2017-03-12 Gerald Pfeifer <gerald@pfeifer.com>
1484 * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
1485 ftp.kernel.org and simplify binutils requirement.
1487 2017-03-11 Gerald Pfeifer <gerald@pfeifer.com>
1489 * doc/invoke.texi (Warning Options): Fix spelling of link-time
1491 (Optimize Options): Ditto. Also remove redundancy.
1493 2017-03-10 David Malcolm <dmalcolm@redhat.com>
1495 PR translation/79848
1496 * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
1498 * ipa-pure-const.c (suggest_attribute): Likewise. Convert _
1499 to G_ to avoid double translation.
1501 2017-03-10 David Malcolm <dmalcolm@redhat.com>
1503 PR translation/79923
1504 * auto-profile.c (get_combined_location): Convert leading
1505 character of diagnostics to lower case and remove trailing period.
1506 (read_profile): Likewise for various diagnostics.
1507 * config/arm/arm.c (arm_option_override): Remove trailing period
1508 from various diagnostics.
1509 * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
1510 (msp430_expand_delay_cycles): Likewise.
1512 2017-03-10 David Malcolm <dmalcolm@redhat.com>
1515 * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
1516 full command-line argument, rather than just "str".
1517 (aarch64_validate_march): Likewise.
1518 (aarch64_validate_mtune): Likewise.
1520 2017-03-10 Bernd Schmidt <bschmidt@redhat.com>
1522 PR rtl-optimization/78911
1523 * lra-assigns.c (must_not_spill_p): New function.
1524 (spill_for): Use it.
1526 2017-03-10 Jakub Jelinek <jakub@redhat.com>
1528 PR tree-optimization/79981
1529 * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
1530 ATOMIC_COMPARE_EXCHANGE ifn result.
1531 (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
1532 IFN_ATOMIC_COMPARE_EXCHANGE.
1534 2017-03-10 David Malcolm <dmalcolm@redhat.com>
1537 * opts.c (parse_sanitizer_options): Add missing question mark to
1538 "did you mean" message.
1540 2017-03-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1542 * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
1544 (VMULEUH_UNS): Likewise.
1545 (VMULOUB_UNS): Likewise.
1546 (VMULOUH_UNS): Likewise.
1547 * config/rs6000/rs6000.c (builtin_function_type): Remove
1548 references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
1550 2017-03-10 David Malcolm <dmalcolm@redhat.com>
1553 * read-rtl-function.c (function_reader::read_rtx_operand): Update
1554 x with result of extra_parsing_for_operand_code_0.
1555 (function_reader::extra_parsing_for_operand_code_0): Convert
1556 return type from void to rtx, returning x. When reading
1557 SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
1558 larger size containing struct block_symbol.
1560 2017-03-10 Segher Boessenkool <segher@kernel.crashing.org>
1562 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
1563 -mfloat128-hardware without -m64.
1565 2017-03-10 Will Schmidt <will_schmidt@vnet.ibm.com>
1568 * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
1569 entries to the case statement that marks unsigned arguments to
1570 overloaded functions.
1572 2017-03-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1574 * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
1575 two typographic errors in the handling of TARGET_UPPER_REGS_DI.
1577 2017-03-10 Pat Haugen <pthaugen@us.ibm.com>
1580 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
1581 TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
1583 2017-03-10 Martin Liska <mliska@suse.cz>
1587 * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
1588 * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
1591 2017-03-10 Olivier Hainque <hainque@adacore.com>
1593 * tree-switch-conversion (array_value_type): Start by resetting
1594 candidate type to it's main variant.
1596 2017-03-10 Jakub Jelinek <jakub@redhat.com>
1598 PR rtl-optimization/79909
1599 * combine.c (try_combine): Use simplify_replace_rtx on individual
1600 CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
1601 of the whole CALL_INSN_FUNCTION_USAGE.
1603 PR tree-optimization/79972
1604 * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
1605 get_range_info on SSA_NAMEs. Formatting fixes.
1607 2017-03-10 Richard Biener <rguenther@suse.de>
1608 Jakub Jelinek <jakub@redhat.com>
1610 PR tree-optimization/77975
1611 * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
1612 edge to be constant.
1613 (get_val_for): For constant x return it. Formatting fix.
1614 (loop_niter_by_eval): Avoid pointless looping if the next iteration
1615 would use the same bases as the current one.
1617 2017-03-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1619 * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
1620 instead of vec_select for V1TImode.
1621 * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
1623 (VSX_LE_128): Add V1TI to this mode iterator.
1624 (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
1625 (*vsx_le_perm_store_<mode>): Likewise.
1626 (pre-reload splitter for VSX stores): Likewise.
1627 (post-reload splitter for VSX stores): Likewise.
1628 (*vsx_xxpermdi2_le_<mode>): Likewise.
1629 (*vsx_lxvd2x2_le_<mode>): Likewise.
1630 (*vsx_stxvd2x2_le_<mode>): Likewise.
1632 2017-03-09 Michael Eager <eager@eagercon.com>
1634 Correct failures with --enable-checking=yes,rtl.
1636 * config/microblaze/microblaze.c (microblaze_expand_shift):
1637 Replace GET_CODE test with CONST_INT_P and INTVAL test with
1638 test for const0_rtx.
1639 * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
1640 lshrsi3_byone): Replace INTVAL with test for const1_rtx.
1642 2017-03-09 Richard Biener <rguenther@suse.de>
1644 PR tree-optimization/79977
1645 * graphite-scop-detection.c (scop_detection::merge_sese):
1646 Handle the case of extra exits to blocks dominating the entry.
1648 2017-03-09 Toma Tabacu <toma.tabacu@imgtec.com>
1650 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
1653 2017-03-09 Vladimir Makarov <vmakarov@redhat.com>
1655 PR rtl-optimization/79949
1656 * lra-constraints.c (process_alt_operands): Check memory when
1657 trying to predict a cycle. Print about the overall increase.
1659 2017-03-09 Richard Biener <rguenther@suse.de>
1662 * gimple-expr.c (useless_type_conversion_p): Preserve
1663 TYPE_SATURATING for fixed-point types.
1665 2017-03-09 Richard Biener <rguenther@suse.de>
1668 * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
1669 alignment of BLKmode params.
1671 2017-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1674 * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
1675 (VALL_NO_V2Q): Likewise.
1677 * config/aarch64/aarch64-simd.md
1678 (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
1680 (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
1681 VALL_NO_V2Q mode iterator.
1682 (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
1684 2017-03-09 Martin Liska <mliska@suse.cz>
1686 PR tree-optimization/79631
1687 * tree-chkp-opt.c (chkp_is_constant_addr): Call
1688 tree_int_cst_sign_bit just for INTEGER constants.
1690 2017-03-09 Martin Liska <mliska@suse.cz>
1694 * toplev.c (process_options): Disable -fcheck-pointer-bounds with
1697 2017-03-09 Marek Polacek <polacek@redhat.com>
1700 * tree.c (inchash::add_expr): Handle TREE_VEC.
1702 2017-03-09 Martin Liska <mliska@suse.cz>
1705 (chkp_narrow_size_and_offset): New function.
1706 (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
1707 (void chkp_parse_bit_field_ref): New function.
1708 (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
1709 (chkp_process_stmt): Use chkp_parse_bit_field_ref.
1711 2017-03-09 Martin Liska <mliska@suse.cz>
1714 * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
1715 (chkp_find_bounds_1): Remove gcc_unreachable.
1717 2017-03-09 Jakub Jelinek <jakub@redhat.com>
1720 * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
1721 BUILT_IN_SYNC*, determine the access type from the size suffix and
1722 always build a MEM_REF with that type. Handle forgotten
1723 BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
1726 * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
1727 _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
1728 _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
1729 _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
1730 _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
1731 _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
1732 _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
1733 _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
1734 _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
1735 _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
1736 _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
1737 _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
1738 _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
1739 _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
1740 _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
1741 _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
1742 _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
1743 _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
1744 _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
1745 _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
1746 _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
1747 _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
1748 _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
1749 _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
1750 _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
1751 _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
1752 _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
1753 _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
1754 _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
1755 _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
1756 definitions outside of __OPTIMIZE__ guarded section.
1759 * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
1760 _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
1761 _mm512_packus_epi32, _mm512_maskz_packus_epi32,
1762 _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
1765 2017-03-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1767 * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
1768 ("vfenez<mode>"): Add missing constraints.
1770 2017-03-08 Martin Sebor <msebor@redhat.com>
1773 * config/nds32/nds32.c (nds32_option_override):
1775 2017-03-08 Jakub Jelinek <jakub@redhat.com>
1778 * gimplify.c (gimplify_omp_for): Replace index var in outer
1779 taskloop statement with an artificial variable and add
1780 OMP_CLAUSE_PRIVATE clause for it.
1782 2017-03-08 Richard Biener <rguenther@suse.de>
1784 PR tree-optimization/79955
1785 * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
1786 for accesses that are completely outside of the variable.
1788 2017-03-08 Andrew Haley <aph@redhat.com>
1790 PR tree-optimization/79943
1791 * tree-ssa-loop-split.c (compute_new_first_bound): When
1792 calculating the new upper bound, (END-BEG) should be added, not
1795 2017-03-08 Jakub Jelinek <jakub@redhat.com>
1797 * config/avr/avr.md (setmemhi): Make sure match_dup
1798 operand number comes before match_scratch.
1800 2017-03-08 Richard Biener <rguenther@suse.de>
1802 PR tree-optimization/79920
1803 * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
1804 with ncopies == 1 to ...
1805 (vect_transform_slp_perm_load): ... here. Properly compute
1806 all element loads by iterating VF times over the group. Do
1807 not handle ncopies (computed in a broken way) in
1808 vect_create_mask_and_perm.
1810 2017-03-08 Jakub Jelinek <jakub@redhat.com>
1813 * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
1814 is a uniform vector, use uniform_vector_p return value instead of
1815 building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
1817 2017-03-07 Marek Polacek <polacek@redhat.com>
1820 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
1821 (alloca_call_type): Likewise.
1823 2017-03-07 Martin Liska <mliska@suse.cz>
1825 * gcov.c (process_args): Put comment to correct location.
1827 2017-03-07 Martin Liska <mliska@suse.cz>
1830 * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
1831 Use array_at_struct_end_p instead of DECL_CHAIN (field).
1832 (chkp_narrow_bounds_for_field): Likewise.
1833 (chkp_parse_array_and_component_ref): Pass one more argument to
1836 2017-03-07 Richard Biener <rguenther@suse.de>
1838 * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
1841 2017-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1843 * config/i386/i386.c (ix86_local_alignment): Align most aggregates
1844 of 16 bytes and more to 16 bytes, not those of 16 bits and more.
1846 2017-03-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1849 * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
1850 to end of description.
1851 (PARAM_MAX_STORES_TO_MERGE): Likewise.
1853 2017-03-07 Jakub Jelinek <jakub@redhat.com>
1855 PR rtl-optimization/79901
1856 * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
1858 (*avx512f_<code><mode>3<mask_name>): ... this.
1859 (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
1860 iterator instead of VI8_AVX2_AVX512BW.
1862 PR rtl-optimization/79901
1863 * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
1864 min/max expander, expand it using expand_vec_cond_expr.
1867 * ubsan.c (ubsan_encode_value): Call mark_addressable on the
1870 2017-03-06 Jakub Jelinek <jakub@redhat.com>
1873 * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
1874 to void * for PCH reasons.
1875 * dwarf2out.c (output_loc_operands, output_die): Cast
1876 v.val_vec.array to unsigned char *.
1878 2017-03-06 John David Anglin <danglin@gcc.gnu.org>
1881 * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
1884 2017-03-06 Vladimir Makarov <vmakarov@redhat.com>
1886 PR rtl-optimization/79571
1887 * lra-constraints.c (process_alt_operands): Calculate static
1888 reject and subtract it from overall when only addresses will be
1891 2017-03-06 Julia Koval <julia.koval@intel.com>
1894 * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
1895 incoming stack boundary to 128 for 64-bit targets.
1897 2017-03-06 Richard Biener <rguenther@suse.de>
1899 PR tree-optimization/79894
1900 * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
1901 to NULL after folding it.
1903 2017-03-06 Richard Biener <rguenther@suse.de>
1905 PR tree-optimization/79824
1906 * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
1907 check disabling peeling for gaps.
1909 2017-03-06 Toma Tabacu <toma.tabacu@imgtec.com>
1911 * doc/sourcebuild.texi (Effective-Target Keywords, Environment
1912 attributes): Document gettimeofday.
1914 2017-03-06 Robin Dapp <rdapp@linux.vnet.ibm.com>
1916 * config/s390/s390.c (s390_option_override_internal): Set
1917 PARAM_MIN_VECT_LOOP_BOUND
1919 2017-03-06 Robin Dapp <rdapp@linux.vnet.ibm.com>
1921 * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
1922 * config/s390/s390.md: Likewise.
1924 2017-03-06 Jakub Jelinek <jakub@redhat.com>
1927 * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
1928 (<avx2_avx512>_perm<mode>): Rename to ...
1929 (avx2_perm<mode>): ... this. Use VI8F_256 iterator instead
1931 (<avx512>_perm<mode>_mask): Rename to ...
1932 (avx512vl_perm<mode>_mask): ... this. Use VI8F_256 iterator instead
1934 (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
1935 (avx2_perm<mode>_1<mask_name): ... this. Use VI8F_256 iterator
1936 instead of VI8F_256_512.
1937 (avx512f_perm<mode>): New define_expand.
1938 (avx512f_perm<mode>_mask): Likewise.
1939 (avx512f_perm<mode>_1<mask_name>): New define_insn.
1940 (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
1942 2017-03-06 Prachi Godbole <prachi.godbole@imgtec.com>
1944 * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
1945 msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
1947 (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
1949 2017-03-06 Martin Liska <mliska@suse.cz>
1952 * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
1953 when having a SSA NAME w/o VAR_DECL assigned to it.
1955 2017-03-06 Prachi Godbole <prachi.godbole@imgtec.com>
1957 * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
1958 msa_dpsub_<su>_d): Fix MODE for vec_select.
1960 2017-03-06 Prachi Godbole <prachi.godbole@imgtec.com>
1962 * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
1964 * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
1966 2017-03-06 Richard Biener <rguenther@suse.de>
1968 * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
1969 * plugin.c (register_plugin_info): Likewise.
1970 * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
1972 2017-03-05 Jakub Jelinek <jakub@redhat.com>
1974 * config/i386/sse.md (sse_storehps, sse_storelps,
1975 avx_<castmode><avxsizesuffix>_<castmode>,
1976 avx512f_<castmode><avxsizesuffix>_<castmode>,
1977 avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
1978 in condition that at least one operand is not a MEM.
1980 2017-03-03 Jakub Jelinek <jakub@redhat.com>
1983 * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
1984 ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
1986 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
1987 gimple_call_nothrow_p flag based on whether original builtin can throw.
1988 If it can, emit following stmts on the fallthrough edge.
1989 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
1990 don't create new bb if inserting just debug stmts on the edge, try to
1991 insert them on the fallthru bb or just reset debug stmts.
1993 2017-03-03 Segher Boesssenkool <segher@kernel.crashing.org>
1996 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
1997 restore recog_data (including the operand rtxes inside it) around
1998 the call to get_insn_template.
2000 2017-03-03 Martin Sebor <msebor@redhat.com>
2002 PR tree-optimization/79699
2003 * context.c (context::~context): Free MPFR caches to avoid
2004 a memory leak on program exit.
2006 2017-03-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2008 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
2009 Use wide_int::ulow () instead of .elt (0).
2011 2017-03-03 Uros Bizjak <ubizjak@gmail.com>
2013 * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
2014 (*pushxf): Limit oF constraint to 32bit targets and add oC
2015 constraint for 64bit targets.
2016 (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
2017 (*pushdf): Change rmF constraint to rmC.
2019 2017-03-03 Martin Liska <mliska@suse.cz>
2021 * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
2022 Remove unused variable.
2024 2017-03-03 Jakub Jelinek <jakub@redhat.com>
2027 * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
2028 is a memory operand, increase num_memory.
2029 (ix86_expand_args_builtin): Likewise.
2031 2017-03-03 Jan Hubicka <jh@suse.cz>
2034 * ipa-devirt.c (maybe_record_node): Properly handle
2035 __cxa_pure_virtual visibility.
2037 2017-03-03 Martin Liska <mliska@suse.cz>
2039 PR tree-optimization/79803
2040 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
2042 (pass_loop_prefetch::execute): Disabled optimization if an
2043 assumption about L1 cache size is not met.
2045 2017-03-03 Martin Liska <mliska@suse.cz>
2047 PR rtl-optimization/79574
2048 * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
2049 (hash_scan_set): Likewise.
2050 (dump_hash_table): Likewise.
2051 (hoist_code): Likewise.
2053 2017-03-03 Richard Biener <rguenther@suse.de>
2055 * fixed-value.c (fixed_from_string): Restore use of elt (1)
2056 in place of uhigh ().
2057 (fixed_convert_from_real): Likewise.
2059 2017-03-03 Uros Bizjak <ubizjak@gmail.com>
2062 * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
2064 2017-03-03 Richard Biener <rguenther@suse.de>
2067 * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
2068 TYPE_OVERFLOW_UNDEFINED check.
2070 2017-03-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2072 * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
2074 (vector_ae_<mode>_p): Likewise.
2075 (vector_nez_<mode>_p): Likewise.
2076 (vector_ne_v2di_p): Likewise.
2077 (vector_ae_v2di_p): Likewise.
2078 (vector_ne_<mode>_p): Likewise.
2079 * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
2081 (vsx_tsqrt<mode>2_fe): Likewise.
2083 2017-03-02 Uros Bizjak <ubizjak@gmail.com>
2086 * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
2088 2017-03-02 Jakub Jelinek <jakub@redhat.com>
2090 PR rtl-optimization/79780
2091 * cprop.c (one_cprop_pass): When second and further conditional trap
2092 in a single basic block is turned into an unconditional trap, turn it
2093 into a deleted note to avoid RTL verification failures.
2095 2017-03-02 Richard Biener <rguenther@suse.de>
2097 * fold-const.c (const_binop): Use ulow () instead of elt (0).
2099 2017-03-02 Richard Biener <rguenther@suse.de>
2101 PR tree-optimization/79345
2103 * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
2104 param and abort the walk, returning -1 if it is hit.
2105 (walk_aliased_vdefs): Take a limit param and pass it on.
2106 * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
2107 defaulting to 0 and return a signed int.
2108 * tree-ssa-uninit.c (struct check_defs_data): New struct.
2109 (check_defs): New helper.
2110 (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
2111 about uninitialized memory.
2112 * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
2113 bogus uninitialized warning.
2114 (fixed_convert_from_real): Likewise.
2116 2017-03-02 Bin Cheng <bin.cheng@arm.com>
2118 PR tree-optimization/66768
2119 * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
2120 iv_use if base object can't be determined.
2122 2017-03-02 Jakub Jelinek <jakub@redhat.com>
2124 PR tree-optimization/79345
2125 * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
2126 * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
2127 (get_pattern_stats): Initialize it.
2128 * genemit.c (gen_expand): Verify match_scratch numbers come after
2129 match_operand/match_dup numbers.
2130 * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
2131 match_scratch numbers.
2132 * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
2134 * config/s390/s390.md (trunctdsd2): Likewise.
2136 2017-03-02 Richard Biener <rguenther@suse.de>
2138 * wide-int.h (wide_int_storage::operator=): Implement in terms
2141 2017-03-02 Richard Biener <rguenther@suse.de>
2143 PR tree-optimization/79777
2144 * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
2145 the to insert expression to sth existing.
2147 2017-03-01 Martin Sebor <msebor@redhat.com>
2150 * gimple-ssa-sprintf.c
2151 (directive::known_width_and_precision): New function.
2152 (format_integer): Use it.
2153 (get_mpfr_format_length): Consider the full range of precision
2154 when computing %g output with the # flag. Set the likely byte
2155 count to 3 rather than 1 when precision is indeterminate.
2156 (format_floating): Correct the lower bound of precision.
2158 2017-03-01 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2160 * doc/invoke.texi: Document default code model for 64-bit Linux.
2162 2017-03-01 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2165 * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
2166 udiv rather than div since input pattern is unsigned.
2168 2017-03-01 Uros Bizjak <ubizjak@gmail.com>
2170 * config/i386/i386.c (print_reg): Warn for values of
2171 unsupported size in integer register.
2173 2017-03-01 Michael Meissner <meissner@linux.vnet.ibm.com>
2176 * config/rs6000/predicates.md (current_file_function_operand): Do
2177 not allow self calls to be local if the function is replaceable.
2179 2017-03-01 Kelvin Nilsen <kelvin@gcc.gnu.org>
2182 * config/rs6000/altivec.h (vec_ctz and others): Change the
2183 preprocessor macro that controls conditional compilation from
2184 _ARCH_PWR9 to __POWER9_VECTOR__.
2185 (vec_all_ne): Change parameterization of __altivec_scalar_pred
2186 macro expansion under preprocessor #ifdef __POWER9_VECTOR__
2187 control (instead of _ARCH_PWR9 control) so that template
2188 definition uses power9-specific function.
2189 (vec_any_eq): Likewise.
2190 (vec_all_ne): Change macro definition to use a power9-specific
2191 expansion under #ifdef __POWER9_VECTOR__ control (instead of
2192 _ARCH_PWR9 control).
2193 (vec_any_eq) Likewise.
2194 * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
2195 expansion for CMPNEF to remove support for xvcmpnesp instruction.
2196 (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
2197 support for xvcmpnedp instruction.
2198 (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
2199 macro expansion so that Power9 implementation of vec_all_ne does
2200 not use the AltiVec predicate framework.
2201 (VCMPNEH_P): Likewise.
2202 (VCMPNEW_P): Likewise.
2203 (VCMPNED_P): Likewise.
2204 (VCMPNEFP_P): Likewise.
2205 (VCMPNEDP_P): Likewise.
2206 (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
2207 implementation of vec_any_eq to not use AltiVec predicate
2209 (VCMPAEH_P): Likewise.
2210 (VCMPAEW_P): Likewise.
2211 (VCMPAED_P): Likewise.
2212 (VCMPAEFP_P): Likewise.
2213 (VCMPAEDP_P): Likewise.
2214 (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
2215 BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
2216 not use the AltiVec predicate framework.
2217 (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
2218 of vec_any_eq to not use AltiVec predicate framework.
2219 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
2220 support for predefined __POWER9_VECTOR__ macro to indicate that
2221 Power9 instruction selection is enabled.
2222 (altivec_overloaded_builtins): Remove extraneous
2223 ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
2224 function argument types RS6000_BTI_bool_V16QI and
2225 RS6000_BTI_bool_V16QI. Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
2226 entry for overloaded function argument types RS6000_BTI_bool_V4SI
2227 andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB. Remove
2228 two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
2229 P9V_BUILTIN_CMPNEF to force use of instructions not specific to
2230 Power9 for implementations of vec_cmpne. Change the signature for
2231 all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
2232 (representing vec_all_ne) to remove the previously described first
2233 argument of type RS6000_BTI_INTSI, as this was an artifact of
2234 reliance on the AltiVec predicate framework, which is no longer
2235 used in the implementation of these functions. Add
2236 P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
2237 function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
2238 since, unlike the AltiVec predicate framework implementation, we
2239 do not share function descriptors between vec_alle and vec_anyeq.
2240 (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
2241 set of modes that receive special treatment even when
2242 TARGET_P9_VECTOR is true. The special treatment emits code that
2243 does not depend on Power9 instructions.
2244 * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
2245 define_expand to not rely on AltiVec predicate framework.
2246 (vector_ae_<mode>p): New define_expand to represent vec_any_eq
2248 (vector_ne_v2di_p): Change this define_expand to not rely on
2249 AltiVec predicate framework.
2250 (vector_ae_v2di_p): New define_expand to represent vec_any_eq
2252 (vector_ne_<mode>_p): Change this define_expand to not rely on
2253 AltiVec predicate framework.
2254 (vector_ae_<mode>p): New define_expand to represent vec_any_eq
2256 * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
2257 (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
2258 define_insn pattern.
2259 (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
2260 define_insn pattern because the xvcmpne<VSs>. instruction is not
2262 (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
2263 instruction is not supported.
2265 2017-03-01 Jakub Jelinek <jakub@redhat.com>
2267 * config/nvptx/nvptx.c: Include intl.h.
2269 2017-03-01 Martin Jambor <mjambor@suse.cz>
2272 * ipa-prop.h (ipa_bits): Removed field known.
2273 (ipa_jump_func): Removed field vr_known. Changed fields bits and m_vr
2274 to pointers. Adjusted their comments to warn about their sharing.
2275 (ipcp_transformation_summary): Change bits to a vector of pointers.
2276 (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
2277 (ipa_get_ipa_bits_for_value): Declare.
2278 * tree-vrp.h (value_range): Mark as GTY((for_user)).
2279 * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
2280 (ipa_bits_hash_table): Likewise.
2281 (ipa_vr_ggc_hash_traits): Likewise.
2282 (ipa_vr_hash_table): Likewise.
2283 (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
2284 being pointers and vr_known being removed.
2285 (ipa_set_jf_unknown): Likewise.
2286 (ipa_get_ipa_bits_for_value): New function.
2287 (ipa_set_jfunc_bits): Likewise.
2288 (ipa_get_value_range): New overloaded functions.
2289 (ipa_set_jfunc_vr): Likewise.
2290 (ipa_compute_jump_functions_for_edge): Use the above functions to
2291 construct bits and vr parts of jump functions.
2292 (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
2293 ipa_bits_hash_table and ipa_vr_hash_table if they do not already
2295 (ipcp_grow_transformations_if_necessary): Also allocate
2296 ipa_bits_hash_table and ipa_vr_hash_table if they do not already
2298 (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
2299 them. Fix too long lines.
2300 (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
2301 vr_known being removed.
2302 (ipa_read_jump_function): Use new setter functions to construct bits
2303 and vr parts of jump functions or set them to NULL.
2304 (write_ipcp_transformation_info): Adjust for bits being pointers.
2305 (read_ipcp_transformation_info): Likewise.
2306 (ipcp_update_bits): Likewise. Fix excessively long lines a trailing
2308 Include gt-ipa-prop.h.
2309 * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
2311 (ipcp_store_bits_results): Likewise.
2312 (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
2313 Do not write to existing jump functions but use a temporary instead.
2315 2017-03-01 Jakub Jelinek <jakub@redhat.com>
2318 * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
2319 attempt to use its first operand as BIT_FIELD_REF base.
2321 2017-03-01 Richard Biener <rguenther@suse.de>
2324 * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
2325 interpolating formula in wrapping arithmetic.
2326 (chrec_apply): Convert chrec_evaluate return value to wanted type.
2328 2017-03-01 Jakub Jelinek <jakub@redhat.com>
2330 PR tree-optimization/79734
2331 * tree-vect-generic.c (expand_vector_condition): Optimize
2332 AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
2333 Handle VEC_COND_EXPR where comparison has different inner width from
2336 2017-02-28 Sandra Loosemore <sandra@codesourcery.com>
2338 * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
2339 markup, and similar issues. Remove @opindex entries for things
2340 that aren't options. Add missing -mmpy-option entries.
2342 2017-02-28 Jakub Jelinek <jakub@redhat.com>
2344 PR tree-optimization/79737
2345 * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
2346 a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
2347 tmpbuf[byte_size - 1]. Call natice_encode_expr with byte_size - 1
2348 instead of byte_size. Formatting fix.
2349 (shift_bytes_in_array_right): Formatting fix.
2351 2017-02-28 Eric Botcazou <ebotcazou@adacore.com>
2354 * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
2355 condition on optimize for the leaf function test.
2357 2017-02-28 Martin Liska <mliska@suse.cz>
2360 * read-rtl-function.c (function_reader::handle_unknown_directive):
2361 Bail out when one uses -flto.
2363 2017-02-28 Martin Liska <mliska@suse.cz>
2365 * common.opt: Replace space with tabular for options of <number>
2367 * config/i386/i386.opt: Show <number> value for
2368 -mlarge-data-threshold.
2369 * opts.c (print_filtered_help): Do not display number in hexadecimal
2372 2017-02-28 Martin Liska <mliska@suse.cz>
2374 * common.opt: Fix --help=option -Q for options which are of
2377 2017-02-28 Uros Bizjak <ubizjak@gmail.com>
2379 * config/i386/i386.c (print_reg): Error out for values
2380 of 8-bit size in invalid integer register.
2382 2017-02-28 Martin Sebor <msebor@redhat.com>
2384 PR tree-optimization/79691
2385 * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
2387 2017-02-28 Jakub Jelinek <jakub@redhat.com>
2390 * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
2391 gcc_unreachable with output_operand_lossage.
2393 2017-02-28 Richard Biener <rguenther@suse.de>
2395 PR tree-optimization/79740
2396 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
2398 (visit_nary_op): Insert the nary into the hashtable if we
2399 pattern-matched sth.
2400 * tree-ssa-pre.c (eliminate_insert): Robustify.
2402 2017-02-28 Richard Biener <rguenther@suse.de>
2405 * fold-const.c (decode_field_reference): Reject out-of-bound
2408 2017-02-28 Jakub Jelinek <jakub@redhat.com>
2410 * config/i386/i386.c: Include intl.h.
2411 (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
2412 instead of just cond ? "..." : "...".
2413 * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
2414 * coverage.c (read_counts_file): Likewise.
2415 * omp-offload.c: Include intl.h.
2416 (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
2417 of just cond ? "..." : "...".
2418 * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
2419 of just cond ? "..." : "...".
2421 2017-02-28 Richard Earnshaw <rearnsha@arm.com>
2424 * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
2426 * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
2427 'tune for' CPU name.
2428 * config/arm/arm-cpu-data.h: Regenerated.
2430 2017-02-28 Richard Biener <rguenther@suse.de>
2432 PR tree-optimization/79732
2433 * tree-inline.c (expand_call_inline): Do not shadow var.
2435 2017-02-28 Richard Biener <rguenther@suse.de>
2437 PR tree-optimization/79723
2438 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
2439 address-space properly.
2441 2017-02-28 Thomas Schwinge <thomas@codesourcery.com>
2443 * doc/optinfo.texi (Optimization groups): Fix option used for
2445 * doc/invoke.texi (-fopt-info): Document "omp".
2446 * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
2447 (OPTGROUP_ALL): Add OPTGROUP_OMP.
2448 * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
2449 * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
2450 * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
2452 * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP. Adjust
2454 * dumpfile.c (optgroup_options): Instead of "openmp", associate
2455 OPTGROUP_OMP with "omp".
2457 2017-02-27 Pat Haugen <pthaugen@us.ibm.com>
2460 * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
2461 for arithmetic shift of unsigned V2DI.
2463 2017-02-27 Claudiu Zissulescu <claziss@synopsys.com>
2465 * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
2466 arc/linux.h headers.
2467 * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
2468 (LINK_SPEC): Likewise.
2469 (ARC_TLS_EXTRA_START_SPEC): Likewise.
2470 (EXTRA_SPECS): Likewise.
2471 (STARTFILE_SPEC): Likewise.
2472 (ENDFILE_SPEC): Likewise.
2473 (LIB_SPEC): Likewise.
2474 (TARGET_SDATA_DEFAULT): Likewise.
2475 (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
2476 (MULTILIB_DEFAULTS): Likewise.
2477 (DWARF2_UNWIND_INFO): Likewise.
2478 * config/arc/big.h: New file.
2479 * config/arc/elf.h: Likewise.
2480 * config/arc/linux.h: Likewise.
2481 * config/arc/t-uClibc: Remove.
2483 2017-02-27 Bin Cheng <bin.cheng@arm.com>
2485 PR tree-optimization/77536
2486 * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
2487 (tree_transform_and_unroll_loop): Use above function to compute the
2488 estimated niter of unrolled loop and use it when scaling profile.
2489 Also use count info rather than frequency if it's non-zero.
2490 * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
2491 * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
2492 (vect_transform_loop): Call above function.
2494 2017-02-27 Richard Biener <rguenther@suse.de>
2496 PR tree-optimization/45397
2497 * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
2498 * tree-ssa-sccvn.c (valueized_wider_op): New helper.
2499 (visit_nary_op): Add pattern matching for CSEing sign-changed
2500 or truncated operations with wider ones.
2502 2017-02-27 Richard Biener <rguenther@suse.de>
2504 PR tree-optimization/79690
2505 * tree-vect-stmts.c (vectorizable_store): Use vector type
2506 built from the DR with address-space.
2508 2017-02-26 Gerald Pfeifer <gerald@pfeifer.com>
2510 * doc/invoke.texi (Optimize Options): Refine the description
2511 of asan-use-after-return.
2513 2017-02-25 Alan Modra <amodra@gmail.com>
2515 PR rtl-optimization/79584
2516 * lra-constraints.c (base_to_reg): Reload ad->base, the entire
2517 base, not ad->base_term, the reg within base. Remove assertion
2518 that ad->base == ad->base_term. Replace gen_int_mode using
2519 bogus mode with const0_rtx.
2521 2017-02-25 Jakub Jelinek <jakub@redhat.com>
2524 * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
2525 FMA_EXPR like tcc_binary or tcc_unary.
2527 * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
2530 * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
2532 (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
2533 (output_loc_operands): Handle DW_OP_call_ref and
2534 DW_OP_GNU_variable_value.
2535 (struct variable_value_struct): New type.
2536 (struct variable_value_hasher): Likewise.
2537 (variable_value_hash): New variable.
2538 (string_types): Remove.
2539 (copy_loc_descr): New function.
2540 (add_loc_descr_to_each): Clarify comment. Use copy_loc_descr.
2541 (prepend_loc_descr_to_each): New function.
2542 (add_loc_list): Fix comment typo. Use prepend_loc_descr_to_each
2543 instead of add_loc_descr_to_each if the first argument is single
2544 location list and the second has multiple.
2545 (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
2546 (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
2547 when looking for variable value which doesn't have other location info.
2548 (loc_list_from_tree): Formatting fix.
2549 (gen_array_type_die): Simplify DW_AT_string_length handling.
2550 (adjust_string_types): Remove.
2551 (gen_subprogram_die): Don't call adjust_string_types nor test/set
2552 string_types. Call resolve_variable_values.
2553 (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
2554 (resolve_addr_in_expr): Likewise. Add A argument.
2555 (copy_deref_exprloc): Remove deref argument. Adjust for the
2556 original expression being DW_OP_GNU_variable_value with optionally
2557 DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
2558 optionally after it.
2559 (optimize_string_length): Rework for DW_OP_GNU_variable_value.
2560 (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
2561 callers. Set remove_AT_byte_size if removing DW_AT_string_length.
2562 (variable_value_hasher::hash, variable_value_hasher::equal): New
2564 (resolve_variable_value_in_expr, resolve_variable_value,
2565 resolve_variable_values, note_variable_value_in_expr,
2566 note_variable_value): New functions.
2567 (dwarf2out_early_finish): Call note_variable_value on all toplevel
2570 2017-02-24 Jakub Jelinek <jakub@redhat.com>
2573 * opts.h (handle_generated_option): Add GENERATED_P argument.
2574 * opts-common.c (handle_option): Adjust function comment.
2575 (handle_generated_option): Add GENERATED_P argument, pass it to
2577 (control_warning_option): Pass false to handle_generated_option
2579 * opts.c (maybe_default_option): Pass true to handle_generated_option
2581 * optc-gen.awk: Likewise.
2583 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
2585 * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
2586 a REG, look at the REG it is a SUBREG of.
2587 (splitter for cmpeqsi_t): Ditto.
2589 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
2591 * config/pa/pa.c (pa_combine_instructions): Do not share RTL. Make
2592 the special USEs with the pattern of the insn, not the insn itself.
2594 2017-02-24 Matthew Fortune <matthew.fortune@imgtec.com>
2597 * doc/invoke.texi: Document -mload-store-pairs.
2599 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
2600 Sandra Loosemore <sandra@codesourcery.com>
2602 * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
2603 argument isn't a CONST_INT.
2604 (nios2_alternate_compare_const): Assert op is a CONST_INT.
2605 (nios2_valid_compare_const_p): Assert op is a CONST_INT.
2606 (nios2_validate_compare): Bypass alternate compare logic if *op2
2608 (ldstwm_operation_p): Return false if first_base is not a REG or
2609 if first_offset is not a CONST_INT.
2611 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
2613 * config/cris/cris.md: Use correct operand in a define_peephole2.
2615 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
2617 * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
2619 2017-02-24 Segher Boessenkool <segher@kernel.crashing.org>
2621 * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
2622 this_insn if it is an INSN or JUMP_INSN.
2623 (force_offsettable): Look at base, not at addr.
2624 * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
2625 on things that aren't necessarily CONST_INTs.
2627 2017-02-24 Uros Bizjak <ubizjak@gmail.com>
2629 * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
2630 -mfpmath=sse is the default also for x86-32 targets with SSE2
2631 instruction set when @option{-ffast-math} is enabled
2633 2017-02-24 Jeff Law <law@redhat.com>
2635 PR rtl-optimizatoin/79286
2636 * ira.c (update_equiv_regs): Drop may_trap_p exception to
2639 2017-02-24 Richard Biener <rguenther@suse.de>
2641 PR tree-optimization/79389
2642 * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
2645 2017-02-24 Aldy Hernandez <aldyh@redhat.com>
2647 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
2648 function comment to reflect reality.
2649 (loop_exits_before_overflow): Fix typo in function description.
2651 2017-02-24 Richard Biener <rguenther@suse.de>
2653 PR tree-optimization/79389
2654 * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
2655 properly that a threading opportunity exists. Detect conditional
2656 copy/constant propagation opportunities.
2658 2017-02-23 Eric Botcazou <ebotcazou@adacore.com>
2660 * config/visium/visium.md (type): Add trap.
2661 (b): New mode attribute.
2662 (*btst): Rename into...
2663 (*btst<mode>): ...this and adjust.
2664 (*cbranchsi4_btst_insn): Rename into...
2665 (*cbranch<mode>4_btst_insn): ...this and adjust.
2666 (trap): New define_insn.
2668 2017-02-23 Jakub Jelinek <jakub@redhat.com>
2670 PR tree-optimization/79389
2671 * ifcvt.c (struct noce_if_info): Add rev_cond field.
2672 (noce_reversed_cond_code): New function.
2673 (noce_emit_store_flag): Use rev_cond if non-NULL instead of
2674 reversed_comparison_code. Formatting fix.
2675 (noce_try_store_flag): Test rev_cond != NULL in addition to
2676 reversed_comparison_code.
2677 (noce_try_store_flag_constants): Likewise.
2678 (noce_try_store_flag_mask): Likewise.
2679 (noce_try_addcc): Use rev_cond if non-NULL instead of
2680 reversed_comparison_code.
2681 (noce_try_cmove_arith): Likewise. Formatting fixes.
2682 (noce_try_minmax, noce_try_abs): Clear rev_cond.
2683 (noce_find_if_block): Initialize rev_cond.
2684 (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
2685 instead of false as last argument never attempt to reverse it
2688 2017-02-23 Bin Cheng <bin.cheng@arm.com>
2690 PR tree-optimization/79663
2691 * tree-predcom.c (combine_chains): Process refs in reverse order
2692 only for ZERO length chains, and add explaining comment.
2694 2017-02-23 Jeff Law <law@redhat.com>
2696 PR tree-optimization/79578
2697 * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
2698 in call to operand_equal_p.
2700 2017-01-23 Dominique d'Humieres <dominiq@lps.ens.fr>
2703 * config/i386/cpuid.h: Fix another undefined behavior.
2705 2017-02-23 Richard Biener <rguenther@suse.de>
2707 PR tree-optimization/79683
2708 * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
2709 vector types for data-refs.
2711 2017-02-23 Martin Liska <mliska@suse.cz>
2713 * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
2715 2017-02-23 Jakub Jelinek <jakub@redhat.com>
2718 * internal-fn.c (get_range_pos_neg): Moved to ...
2719 * tree.c (get_range_pos_neg): ... here. No longer static.
2720 * tree.h (get_range_pos_neg): New prototype.
2721 * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
2722 are known to be in between 0 and signed maximum inclusive, try to
2723 expand both unsigned and signed divmod and use the cheaper one from
2726 2017-02-22 Jeff Law <law@redhat.com>
2728 PR tree-optimization/79578
2729 * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
2730 to compare base operands.
2732 2017-02-22 Segher Boessenkool <segher@kernel.crashing.org>
2735 * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
2736 gpc_reg_operand instead of fpr_reg_operand.
2738 2017-02-22 Sameera Deshpande <sameera.deshpande@imgtec.com>
2740 * config/mips/mips.c (mips_return_in_memory): Force FP
2741 vector types to be returned in memory for o32 ABI.
2743 2017-02-22 Jakub Jelinek <jakub@redhat.com>
2745 * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
2746 instead of DW_TAG_member for static data member declarations and don't
2747 set no_linkage_name for static inline data members.
2748 (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
2751 2017-02-22 Martin Liska <mliska@suse.cz>
2753 * doc/invoke.texi: Replace inequality signs with square brackets
2756 2017-02-22 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2758 PR tree-optimization/68644
2759 * gcc.dg/tree-ssa/ivopts-lt-2.c: Skip for powerpc*-*-*.
2761 2017-02-22 Matthew Fortune <matthew.fortune@imgtec.com>
2764 * lra-constraints.c (simplify_operand_subreg): Handle
2765 WORD_REGISTER_OPERATIONS targets.
2767 2017-02-22 Jakub Jelinek <jakub@redhat.com>
2770 * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
2771 and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
2772 elimination by swapping fld*.
2774 2017-02-22 Richard Biener <rguenther@suse.de>
2776 PR tree-optimization/79673
2777 * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
2778 convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
2779 irrelevant address-space qualifiers and avoiding a
2780 ADDR_SPACE_CONVERT_EXPR from fold_convert.
2782 2017-02-22 Richard Biener <rguenther@suse.de>
2784 PR tree-optimization/79666
2785 * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
2786 to not symbolically negate if that may introduce undefined
2789 2017-02-22 Martin Liska <mliska@suse.cz>
2792 * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
2793 * data-streamer-out.c (streamer_write_gcov_count_stream):
2795 * value-prof.c (stream_out_histogram_value): Make assert more
2796 precise based on type of counter.
2798 2017-02-21 Uros Bizjak <ubizjak@gmail.com>
2801 * config/i386/i386.md (standard_x87sse_constant_load splitter):
2802 Use nonimmediate_operand instead of memory_operand for operand 1.
2803 (float-extend standard_x87sse_constant_load splitter): Ditto.
2805 2017-02-21 Jeff Law <law@redhat.com>
2807 PR tree-optimization/79621
2808 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
2809 blocks with edges to themselves.
2811 2017-02-21 Jakub Jelinek <jakub@redhat.com>
2814 * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
2815 is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
2816 Use gimple_call_builtin_p.
2819 * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
2820 on temporarily removed DEBUG_INSNs.
2822 PR tree-optimization/79649
2823 * tree-loop-distribution.c (classify_partition): Give up on
2824 non-generic address space loads/stores.
2826 2017-02-21 Aldy Hernandez <aldyh@redhat.com>
2828 * doc/loop.texi (Loop manipulation): Remove nonexistent
2829 tree_ssa_loop_version from the documentation.
2830 * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
2832 2017-02-21 Jakub Jelinek <jakub@redhat.com>
2835 * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
2836 make_reg_eh_region_note_nothrow_nononlocal on call_insn.
2837 * config/rs6000/rs6000.c: Include except.h.
2838 (rs6000_expand_split_stack_prologue): Call
2839 make_reg_eh_region_note_nothrow_nononlocal on the call insn.
2841 2017-02-21 Martin Jambor <mjambor@suse.cz>
2844 * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
2847 2017-02-21 Martin Jambor <mjambor@suse.cz>
2849 * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
2850 for backward compatibility only.
2851 * doc/invoke.texi (Option Summary): Remove all references to
2854 2017-02-21 Matthew Fortune <matthew.fortune@imgtec.com>
2858 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
2860 * lra-constraints.c (curr_insn_transform): Handle
2861 WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
2863 2017-02-21 Martin Liska <mliska@suse.cz>
2865 * config/i386/i386.opt: Replace -masm-dialect with -masm.
2867 2017-02-21 Thomas Schwinge <thomas@codesourcery.com>
2869 PR translation/79638
2870 * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
2872 2017-02-21 Eric Botcazou <ebotcazou@adacore.com>
2875 * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
2876 (arm_function_ok_for_sibcall): Return false for an indirect call by
2877 descriptor if all the argument registers are used.
2878 (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
2879 alignment of the function.
2881 2017-02-21 Jakub Jelinek <jakub@redhat.com>
2883 PR tree-optimization/61441
2884 * simplify-rtx.c (simplify_const_unary_operation): For
2885 -fsignaling-nans and sNaN operand, return NULL_RTX rather than
2886 the sNaN unmodified.
2888 2017-02-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2890 * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
2891 (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
2892 instead of SYSTEM_HEADER_DIR.
2894 2017-02-20 Gerald Pfeifer <gerald@pfeifer.com>
2895 Martin Liška <mliska@suse.cz>
2897 * doc/invoke.texi (use-after-scope-direct-emission-threshold):
2898 Fix typos and grammar, use active voice, and clarify.
2900 2017-02-20 Marek Polacek <polacek@redhat.com>
2903 * gimplify.c (gimplify_expr): Handle unused *&&L;.
2906 * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
2908 2017-02-20 Jakub Jelinek <jakub@redhat.com>
2911 * config/i386/i386.c (ix86_expand_builtin): Handle
2912 OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
2913 ix86_builtins_isa[fcode].isa as a requirement of those
2914 flags and any other flag in the bitmask.
2915 (ix86_init_mmx_sse_builtins): Use 0 instead of
2916 ~OPTION_MASK_ISA_64BIT as mask.
2917 * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
2918 __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
2919 __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
2920 __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
2922 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
2925 * lra-constraints.c (split_reg): Check requested split mode
2926 is supported by the register.
2928 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
2930 * lra-constraints.c (simplify_operand_subreg): Remove early
2933 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
2936 * lra-constraints.c (curr_insn_transform): Tighten condition
2937 for converting SUBREG reloads from OP_OUT to OP_INOUT.
2939 2017-02-20 Matthew Fortune <matthew.fortune@imgtec.com>
2942 * lra-constraints.c (curr_insn_transform): Handle
2943 WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
2945 2017-02-19 Uros Bizjak <ubizjak@gmail.com>
2948 2016-05-30 Uros Bizjak <ubizjak@gmail.com>
2950 * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
2952 2017-02-19 Jonathan Wakely <jwakely@redhat.com>
2955 * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
2958 2017-02-19 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2960 * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
2963 2017-02-18 Jakub Jelinek <jakub@redhat.com>
2965 * final.c (last_columnnum, override_columnnum): New variables.
2966 (final_start_function): Set last_columnnum, pass it to begin_prologue
2967 hook and pass 0 to dwarf2out_begin_prologue.
2968 (final_scan_insn): Update override_columnnum. Pass last_columnnum
2969 to source_line debug hook.
2970 (notice_source_line): Compute last_columnnum and for debug_column_info
2971 return true on column changes.
2972 * debug.h (struct gcc_debug_hooks): Add column argument to
2973 source_line and begin_prologue hooks.
2974 (debug_nothing_int_charstar_int_bool): Remove prototype.
2975 (debug_nothing_int_int_charstar,
2976 debug_nothing_int_int_charstar_int_bool): New prototypes.
2977 (dwarf2out_begin_prologue): Add column argument.
2978 * debug.c (do_nothing_debug_hooks): Adjust source_line and
2979 begin_prologue hooks.
2980 (debug_nothing_int_charstar_int_bool): Remove.
2981 (debug_nothing_int_int_charstar,
2982 debug_nothing_int_int_charstar_int_bool): New functions.
2983 * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
2984 through to dwarf2out_source_line.
2985 (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
2986 (dwarf2out_source_line): Add column argument, emit it if requested.
2987 * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
2989 * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
2990 * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
2991 * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
2992 through to dwarf2out_begin_prologue.
2993 (vmsdbgout_source_line): Add column argument, pass it through to
2994 dwarf2out_source_line.
2995 * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
2996 dbxout_source_line caller.
2997 (dbxout_source_line): Add column argument.
2999 * common.opt (gno-column-info, gcolumn-info): New options.
3000 * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
3001 (check_die): Also test for multiple DW_AT_decl_column attributes.
3002 (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
3003 DW_AT_decl_column if requested.
3004 (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
3006 (gen_variable_die): Likewise.
3007 (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
3008 * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
3011 * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
3012 * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
3013 (ix86_handle_option): Handle OPT_m3dnowa.
3014 * doc/invoke.texi (-m3dnowa): Document.
3015 * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
3016 -m3dnowa instead of -m3dnow -march=athlon.
3019 * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
3020 instead of gcc_assert for K, r and R code checks. Formatting fixes.
3022 2017-02-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3025 * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
3026 support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
3027 * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
3028 generator for vsx_xxpermdi_<mode>_be.
3029 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
3030 force big-endian semantics.
3031 (vsx_xxpermdi_<mode>_be): New define_expand with same
3032 implementation as previous version of vsx_xxpermdi_<mode>.
3034 2017-02-17 Jakub Jelinek <jakub@redhat.com>
3036 PR tree-optimization/79327
3037 * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
3038 variable, its initialization and use.
3040 2017-02-17 Julia Koval <julia.koval@intel.com>
3042 * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
3043 (OPTION_MASK_ISA_PKU_UNSET): New.
3044 (ix86_handle_option): Handle -mrdpid.
3045 * config/i386/cpuid.h (bit_RDPID): New.
3046 * config/i386/driver-i386.c (host_detect_local_cpu):
3047 Detect RDPID feature.
3048 * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
3049 * config/i386/i386-c.c (ix86_target_macros_internal):
3051 * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
3052 (ix86_valid_target_attribute_inner_p): Add "rdpid".
3053 (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
3054 * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
3055 * config/i386/i386.md (define_insn "rdpid"): New.
3056 * config/i386/i386.opt Add -mrdpid.
3057 * config/i386/immintrin.h (_rdpid_u32): New.
3059 2017-02-17 Vladimir Makarov <vmakarov@redhat.com>
3061 PR rtl-optimization/79541
3062 * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
3063 instead of transforming it into USE.
3065 2017-02-17 Segher Boessenkool <segher@kernel.crashing.org>
3067 * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
3068 If HONOR_SNANS (SFmode) force the input to a register.
3069 (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
3070 (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
3071 an frsp or similar insn.
3073 2017-02-17 Martin Liska <mliska@suse.cz>
3075 PR rtl-optimization/79577
3076 * params.def (selsched-max-sched-times): Increase minimum to 1.
3078 2017-02-17 Martin Liska <mliska@suse.cz>
3080 PR rtl-optimization/79574
3081 * gcse.c (want_to_gcse_p): Prevent integer overflow.
3083 2017-02-17 Martin Liska <mliska@suse.cz>
3085 PR tree-optimization/79529
3086 * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
3087 ssa_defined_default_def_p to handle cases which are implicitly
3089 * tree-ssa.c (ssa_defined_default_def_p): New function.
3090 (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
3091 which are implicitly defined.
3092 * tree-ssa.h (ssa_defined_default_def_p): Declare.
3094 2017-02-17 Richard Biener <rguenther@suse.de>
3097 * params.def (max-ssa-name-query-depth): Limit to 10.
3099 2017-02-17 Richard Biener <rguenther@suse.de>
3101 PR tree-optimization/79552
3102 * tree-ssa-structalias.c (visit_loadstore): Properly verify
3105 2017-02-17 Richard Biener <rguenther@suse.de>
3108 * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
3110 2017-02-17 Marek Polacek <polacek@redhat.com>
3113 * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
3114 (fold_negate_expr): New wrapper.
3116 2017-02-16 Sandra Loosemore <sandra@codesourcery.com>
3118 * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]:
3119 Correct terminology and de-emphasize pre-standard behavior.
3121 2017-02-16 Alan Modra <amodra@gmail.com>
3123 PR rtl-optimization/79286
3124 * ira.c (def_dominates_uses): New function.
3125 (update_equiv_regs): Don't create an equivalence for insns that
3126 may trap where the register def does not dominate the use.
3128 2017-02-16 Vladimir Makarov <vmakarov@redhat.com>
3130 PR rtl-optimization/78127
3131 * lra.c (lra): Call lra_eliminate before finish the loop after
3134 2017-02-16 Richard Biener <rguenther@suse.de>
3136 * graphite.h: Do not include isl/isl_val_gmp.h, instead include
3138 * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
3139 (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
3140 * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
3141 (isl_val_int_from_wi): New function.
3142 (extract_affine_gmp): Rename to ...
3143 (extract_affine_wi): ... this, take a widest_int.
3144 (extract_affine_int): Just wrap extract_affine_wi.
3145 (add_param_constraints): Use isl_val_int_from_wi.
3146 (add_loop_constraints): Likewise, and extract_affine_wi.
3148 2017-02-15 Jeff Law <law@redhat.com>
3151 * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
3152 ira_init_register_move_cost_if_necessary.
3154 2017-02-15 Martin Sebor <msebor@redhat.com>
3157 * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
3158 removed in a prior commit.
3160 2017-02-15 Bin Cheng <bin.cheng@arm.com>
3162 PR tree-optimization/79347
3163 * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
3164 counters during peeling.
3166 2017-02-15 Thomas Schwinge <thomas@codesourcery.com>
3168 * Makefile.in (site.exp): Remove "set ISLVER".
3170 2017-02-15 Jakub Jelinek <jakub@redhat.com>
3173 * real.c (real_from_integer): Call real_convert even for decimal.
3175 2017-02-15 Dominik Vogt <vogt@linux.vnet.ibm.com>
3178 * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
3180 2017-02-14 Andrew Pinski <apinski@cavium.com>
3182 * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
3183 cores and change the partno/implementer to be correct.
3184 (thunderx2t99p1): New core which replaces thunderx2t99 and still has
3185 the 'B" as the implementer.
3186 * config/aarch64/aarch64-tune.md: Regenerate.
3188 2017-02-14 Carl Love <cel@us.ibm.com>
3190 * config/rs6000/rs6000.c: Add case statement entry to make the
3191 xvcvuxdsp built-in argument unsigned.
3192 * config/rs6000/vsx.md: Fix the source and return operand types so they
3193 match the instruction definitions from the ISA document. Fix typo
3194 in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
3197 2017-02-14 Vladimir Makarov <vmakarov@redhat.com>
3200 * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
3201 member early_clobber_alts.
3202 * lra-lives.c (reg_early_clobber_p): New.
3203 (process_bb_lives): Use it.
3204 * lra.c (new_insn_reg): New arg early_clobber_alts. Use it.
3205 (debug_operand_data): Initialize early_clobber_alts.
3206 (setup_operand_alternative): Set up early_clobber_alts.
3207 (collect_non_operand_hard_regs): Ditto. Pass early clobber
3208 alternatives to new_insn_reg.
3209 (add_regs_to_insn_regno_info): Add arg early_clobber_alts. Use
3211 (lra_update_insn_regno_info): Pass the new arg.
3213 2017-02-14 Jakub Jelinek <jakub@redhat.com>
3216 * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
3217 (new_oacc_loop_raw): Don't clear already cleared fields.
3220 * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
3221 _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
3222 _mm512_prefetch_i64gather_ps): New inline functions and macros.
3224 2017-02-14 Uros Bizjak <ubizjak@gmail.com>
3227 * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
3229 2017-02-14 H.J. Lu <hongjiu.lu@intel.com>
3232 * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
3233 the extra instruction to the right place to store 128-bit constant
3236 2017-02-14 Martin Sebor <msebor@redhat.com>
3239 * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
3240 warning for strings of unknown length.
3242 2017-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3244 * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
3246 2017-02-14 Jeff Law <law@redhat.com>
3249 * ira-costs.c (scan_one_insn): Initialize register move costs
3250 for pseudos seen in USE/CLOBBER insns.
3252 PR tree-optimization/79095
3253 * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
3254 if the numerator has the range ~[0,0] make the resultant range ~[0,0].
3255 (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
3256 if the operands are known to be not equal, then the resulting range
3258 (intersect_ranges): If the new range is ~[0,0] and the old range is
3259 wide, then prefer ~[0,0].
3260 * tree-vrp.c (overflow_comparison_p_1): New function.
3261 (overflow_comparison_p): New function.
3262 * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
3263 if NAME is used in an overflow test.
3264 (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
3265 overflow check that can be expressed as an equality test, then adjust
3266 ops to be that equality test.
3268 2017-02-14 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3270 * config/s390/s390-builtin-types.def: Remove flags argument.
3271 * config/s390/s390.c (s390_init_builtins): Likewise.
3273 2017-02-14 Martin Liska <mliska@suse.cz>
3275 * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
3276 vector. Fix trailing white spaces.
3278 2017-02-14 James Greenhalgh <james.greenhalgh@arm.com>
3280 * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
3283 2017-02-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3285 PR rtl-optimization/68664
3286 * config/arm/arm.c (arm_sched_can_speculate_insn):
3287 New function. Declare prototype.
3288 (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
3290 2017-02-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3292 PR rtl-optimization/68664
3293 * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
3295 (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
3297 2017-02-14 Amit Pawar <amit.pawar@amd.com>
3299 * config/i386/i386.c (znver1_cost): Fix the alignment for function and
3300 max skip bytes for function, loop and jump.
3302 2017-02-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3304 * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
3305 ABS_EXPR for gimple dump.
3307 2017-02-14 Jakub Jelinek <jakub@redhat.com>
3310 * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
3312 PR tree-optimization/79408
3313 * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
3314 case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
3315 (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
3316 also if rhs1 is INTEGER_CST.
3318 2017-02-14 Richard Biener <rguenther@suse.de>
3321 * tree-into-ssa.c (insert_phi_nodes): When the function can
3322 have abnormal edges rewrite SSA names with broken use-def
3323 dominance out of SSA and register them for PHI insertion.
3325 2017-02-13 Martin Sebor <msebor@redhat.com>
3328 * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
3329 clearing info.nowrite flag when snprintf size argument is a range.
3331 2017-02-13 Jakub Jelinek <jakub@redhat.com>
3333 * cprop.c (cprop_jump): Add missing space in string literal.
3334 * tree-ssa-structalias.c (rewrite_constraints): Likewise.
3335 (get_constraint_for_component_ref): Likewise.
3336 * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
3337 * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
3338 * lra-constraints.c (process_alt_operands): Likewise.
3339 * ipa-inline.c (inline_small_functions): Likewise.
3340 * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
3341 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3342 * trans-mem.c (diagnose_tm_1_op): Likewise.
3343 * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
3344 (grid_parallel_clauses_gridifiable): Likewise.
3346 * config/nvptx/mkoffload.c (process): Add space in between
3349 * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
3350 "MOD4_SSE_REGS" and "ALL_REGS".
3352 * spellcheck.c (test_data): Add , in between "foo" and "food".
3354 2017-02-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3357 * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
3358 boundary crossing check and subsequent code generation agree.
3360 2017-02-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3362 * config/aarch64/aarch64.c (has_memory_op): Delete.
3363 (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
3366 2017-02-13 Jakub Jelinek <jakub@redhat.com>
3368 PR rtl-optimization/79388
3369 PR rtl-optimization/79450
3370 * combine.c (distribute_notes): When removing TEM_INSN for which
3371 corresponding dest has last value recorded, invalidate that last
3374 2017-02-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3376 * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
3377 of explicit '@'. Add missing assembly comment marker on branch costs
3380 2017-02-13 Nathan Sidwell <nathan@acm.org>
3382 * gengtype-lex.l (<in_struct>): Add '/'.
3384 2017-02-13 Martin Liska <mliska@suse.cz>
3387 * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
3389 2017-02-13 Richard Biener <rguenther@suse.de>
3391 * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
3393 * configure: Re-generate.
3394 * config.in: Likewise.
3395 * graphite-dependences.c: Simplify as if
3396 HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
3397 * graphite-isl-ast-to-gimple.c: Likewise.
3398 * graphite-optimize-isl.c: Likewise.
3399 * graphite-poly.c: Likewise.
3400 * graphite-sese-to-poly.c: Likewise.
3401 * graphite.h: Likewise.
3402 * toplev.c: Include isl/version.h and use isl_version () for
3403 printing the ISL version.
3404 * doc/install.texi: Update ISL requirement.
3406 2017-02-12 Gerald Pfeifer <gerald@pfeifer.com>
3408 * doc/standards.texi (Standards): Update reference to
3411 2017-02-12 Gerald Pfeifer <gerald@pfeifer.com>
3413 * doc/extend.texi (Named Address Spaces): sourceware.org now
3415 * doc/install.texi (Binaries): Ditto.
3418 2017-02-11 Sandra Loosemore <sandra@codesourcery.com>
3420 * doc/cpp.texi: Replace "stringify"/"stringification" with C
3421 standard terminology "stringize"/"stringizing" throughout.
3422 * doc/cppinternals.texi: Likewise.
3424 2017-02-11 Sandra Loosemore <sandra@codesourcery.com>
3426 * doc/extend.texi: Fix some spelling mistakes and typos.
3427 * doc/invoke.texi: Likewise.
3429 2017-02-11 Jan Hubicka <hubicka@ucw.cz>
3432 * params.def (inline-min-speedup) Change from 10 to 8.
3434 2017-02-11 Jakub Jelinek <jakub@redhat.com>
3436 * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
3439 2017-02-11 Jan Hubicka <hubicka@ucw.cz>
3442 * ipa-inline-analysis.c (get_minimal_bb): New function.
3443 (record_modified): Use it.
3444 (remap_edge_change_prob): Handle also ancestor functions.
3446 2017-02-11 Gerald Pfeifer <gerald@pfeifer.com>
3448 * doc/contrib.texi (Contributors): Remove broken link into
3449 the Mauve CVS repository.
3451 2017-02-11 Jakub Jelinek <jakub@redhat.com>
3454 * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
3455 result computation whenever lhs doesn't have vector mode, not
3456 just when it has BLKmode.
3458 2017-02-10 Gerald Pfeifer <gerald@pfeifer.com>
3460 * doc/makefile.texi (profiledbootstrap): Refer to the
3461 installation instructions only in textual form.
3463 2017-02-10 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3466 * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
3468 2017-02-10 Gerald Pfeifer <gerald@pfeifer.com>
3470 * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
3471 (Specific): Update mingw-w64 reference.
3473 (Specific): Remove broken link to Renesas RX processor.
3475 2017-02-10 Richard Biener <rguenther@suse.de>
3477 * toplev.c (process_options): Do not mention obsolete graphite
3478 options when printing sorry message about missing graphite support.
3479 Mention -floop-nest-optimize.
3481 2017-02-10 Christophe Lyon <christophe.lyon@linaro.org>
3483 * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
3484 (vtst_p16): Likewise.
3485 (vtstq_p8): Likewise.
3486 (vtstq_p16): Likewise.
3488 (vtstq_p64): Likewise.
3489 * config/arm/arm_neon.h (vgetq_lane_p64): New.
3490 (vset_lane_p64): New.
3491 (vsetq_lane_p64): New.
3493 2017-02-10 Jakub Jelinek <jakub@redhat.com>
3495 PR tree-optimization/79411
3496 * tree-ssa-reassoc.c (is_reassociable_op): Return false if
3497 stmt operands are SSA_NAMEs used in abnormal phis.
3498 (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
3501 2017-02-09 Jan Hubicka <hubicka@ucw.cz>
3504 * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
3507 2017-02-09 Jan Hubicka <hubicka@ucw.cz>
3509 * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
3511 2017-02-09 Jakub Jelinek <jakub@redhat.com>
3513 * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
3517 * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
3518 not arbitrary TREE_CONSTANT.
3521 * gimplify.c (gimplify_adjust_omp_clauses): Ignore
3522 "omp declare target link" attribute unless is_global_var.
3523 * omp-offload.c (find_link_var_op): Likewise.
3525 2017-02-09 Nathan Sidwell <nathan@codesourcery.com>
3526 Chung-Lin Tang <cltang@codesourcery.com>
3528 * gimplify.c (gimplify_scan_omp_clauses): No special handling for
3530 (gimplify_adjust_omp_clauses): Don't delete TILE.
3531 (gimplify_omp_for): Deal with TILE.
3532 * internal-fn.c (expand_GOACC_TILE): New function.
3533 * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
3535 * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
3536 (expand_oacc_collapse_init): Add LOC paramter. Initialize tile
3538 (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
3539 avoid DIV for outermost collapse var.
3540 (expand_oacc_for): Insert tile element loop as needed. Adjust.
3541 Remove out of date comments, fix whitespace.
3542 * omp-general.c (omp_extract_for_data): Deal with tiling.
3543 * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
3544 adjust OLF_DIM_BASE value.
3545 (struct omp_for_data): Add tiling field.
3546 * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
3547 (lower_oacc_head_mark): Add OLF_TILE as appropriate. Ensure 2 levels
3548 for auto loops. Remove default auto determining, moved to
3549 oacc_loop_fixed_partitions.
3550 * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
3551 stmts, add e_mask field.
3552 (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
3553 (oacc_thread_numbers): Use oacc_dim_call.
3554 (oacc_xform_tile): New.
3555 (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
3556 (finish_oacc_loop): Adjust for ifns vector.
3557 (oacc_loop_discover_walk): Append loop abstraction sites to list,
3558 add case for GOACC_TILE fns.
3559 (oacc_loop_xform_loop): Delete.
3560 (oacc_loop_process): Iterate over call list directly, and add
3561 handling for GOACC_TILE fns.
3562 (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
3564 (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
3565 vector partitioning to outer loops. Assign 2 partitions to loops
3566 when available. Add TILE handling.
3567 (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
3568 (execite_oacc_device_lower): Process GOACC_TILE fns, ignore unknown specs.
3569 * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
3570 * tree.c (omp_clause_num_ops): Adjust TILE ops.
3571 * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
3573 2017-02-09 Gerald Pfeifer <gerald@pfeifer.com>
3575 * configure.ac (ACX_BUGURL): Update.
3576 * configure: Regenerate.
3578 2017-02-09 Richard Biener <rguenther@suse.de>
3580 PR tree-optimization/69823
3581 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3582 Properly enumerate all BBs in the region. Use auto_vec/auto_bitmap.
3584 2017-02-09 Andrew Burgess <andrew.burgess@embecosm.com>
3586 * config/arc/arc-c.def: Add __NPS400__ definition.
3587 * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
3588 (TARGET_NPS400): Define.
3590 2017-02-09 Andrew Burgess <andrew.burgess@embecosm.com>
3592 * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
3594 (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
3596 (arc_cpu_types): Fill the arch_info field with a pointer into the
3597 arc_arch_types table.
3598 (arc_selected_cpu): Declare.
3599 * config/arc/arc.c (arc_selected_cpu): Make global.
3600 (arc_selected_arch): Delete.
3601 (arc_base_cpu): Delete.
3602 (arc_override_options): Remove references to deleted variables,
3603 update access to arch information.
3604 (ARC_OPT): Update access to arch information.
3605 (ARC_OPTX): Likewise.
3606 * config/arc/arc.h (arc_base_cpu): Remove declaration.
3607 (TARGET_ARC600): Update access to arch information.
3608 (TARGET_ARC601): Likewise.
3609 (TARGET_ARC700): Likewise.
3610 (TARGET_EM): Likewise.
3611 (TARGET_HS): Likewise.
3612 * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
3615 2017-02-08 Pat Haugen <pthaugen@us.ibm.com>
3618 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
3619 condition/operands for integer GE/LE/GEU/LEU operations.
3621 2017-02-08 Segher Boessenkool <segher@kernel.crashing.org>
3623 PR translation/79397
3624 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
3627 2017-02-08 Martin Jambor <mjambor@suse.cz>
3630 * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
3631 whether allocation happened.
3632 (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
3633 nothing was allocated.
3635 2017-02-08 Jakub Jelinek <jakub@redhat.com>
3637 PR tree-optimization/79408
3638 * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
3639 constant, but SSA_NAME with a known integer range, use the minimum
3640 of that range instead of op1 to determine if modulo can be replaced
3641 with its first operand.
3643 2016-02-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3645 * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
3647 2017-02-08 Richard Biener <rguenther@suse.de>
3649 PR tree-optimization/71824
3650 * graphite-scop-detection.c (scop_detection::build_scop_breadth):
3651 Check all loops contained in the merged region.
3653 2017-02-07 Andrew Pinski <apinski@cavium.com>
3655 * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
3657 2017-02-07 Andrew Pinski <apinski@cavium.com>
3659 * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
3660 (thunderxt88): Likewise.
3661 (thunderxt81): Disable LSE and change v8.1 to v8.
3662 (thunderxt83): Likewise.
3664 2017-02-07 Jakub Jelinek <jakub@redhat.com>
3665 Richard Biener <rguenther@suse.de>
3668 * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
3669 type from int to size_t.
3670 * ira-costs.c (struct_costs_size): Change type from int to size_t.
3672 2017-02-07 Jakub Jelinek <jakub@redhat.com>
3674 PR rtl-optimization/79386
3675 * cprop.c (bypass_conditional_jumps): Initialize
3676 bypass_last_basic_block already before splitting bbs after
3677 unconditional traps...
3678 (bypass_conditional_jumps): ... rather than here.
3681 * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
3682 (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
3683 *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
3684 fix -masm=intel patterns.
3686 2017-02-07 Richard Biener <rguenther@suse.de>
3688 PR tree-optimization/79256
3690 * builtins.c (get_object_alignment_2): Use min_align_of_type
3691 to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
3692 and ADJUST_FIELD_ALIGN.
3694 * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
3696 * doc/tm.texi: Regenerate.
3697 * stor-layout.c (layout_decl): Adjust.
3698 (update_alignment_for_field): Likewise.
3699 (place_field): Likewise.
3700 (min_align_of_type): Likewise.
3701 * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
3702 * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
3703 * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
3704 * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
3705 * config/frv/frv.c (frv_adjust_field_align): Likewise.
3706 * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
3707 * config/i386/i386.c (x86_field_alignment): Likewise.
3708 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
3709 * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
3710 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
3711 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
3712 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
3713 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
3717 2017-01-30 Richard Biener <rguenther@suse.de>
3719 PR tree-optimization/79256
3720 * targhooks.c (default_builtin_vector_alignment_reachable): Honor
3721 BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
3724 2017-02-07 Toma Tabacu <toma.tabacu@imgtec.com>
3726 * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
3727 argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
3728 builtins to SImode and emit a zero-extend, if necessary.
3730 2017-02-06 Palmer Dabbelt <palmer@dabbelt.com>
3732 * docs/invoke.texi (RISC-V Options): Alphabetize.
3734 2017-02-06 Palmer Dabbelt <palmer@dabbelt.com>
3736 * doc/invoke.texi (RISC-V Options): Use two spaces to separate
3739 2017-02-06 Palmer Dabbelt <palmer@dabbelt.com>
3741 * config/riscv/riscv.c: New file.
3742 * gcc/common/config/riscv/riscv-common.c: Likewise.
3743 * config.gcc: Likewise.
3744 * config/riscv/constraints.md: Likewise.
3745 * config/riscv/elf.h: Likewise.
3746 * config/riscv/generic.md: Likewise.
3747 * config/riscv/linux.h: Likewise.
3748 * config/riscv/multilib-generator: Likewise.
3749 * config/riscv/peephole.md: Likewise.
3750 * config/riscv/pic.md: Likewise.
3751 * config/riscv/predicates.md: Likewise.
3752 * config/riscv/riscv-builtins.c: Likewise.
3753 * config/riscv/riscv-c.c: Likewise.
3754 * config/riscv/riscv-ftypes.def: Likewise.
3755 * config/riscv/riscv-modes.def: Likewise.
3756 * config/riscv/riscv-opts.h: Likewise.
3757 * config/riscv/riscv-protos.h: Likewise.
3758 * config/riscv/riscv.h: Likewise.
3759 * config/riscv/riscv.md: Likewise.
3760 * config/riscv/riscv.opt: Likewise.
3761 * config/riscv/sync.md: Likewise.
3762 * config/riscv/t-elf-multilib: Likewise.
3763 * config/riscv/t-linux: Likewise.
3764 * config/riscv/t-linux-multilib: Likewise.
3765 * config/riscv/t-riscv: Likewise.
3766 * configure.ac: Likewise.
3767 * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
3768 Waterman as RISC-V maintainers.
3769 * doc/install.texi: Add RISC-V entries.
3770 * doc/invoke.texi: Add RISC-V options section.
3771 * doc/md.texi: Add RISC-V constraints section.
3772 * configure: Regenerated.
3774 2017-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
3777 * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
3778 false values to be constant vectors with all 0 or all 1 bits set.
3779 (vcondu<mode><mode>): Likewise.
3780 * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
3782 (fpmask_comparison_operator): Update comment.
3783 (vecint_comparison_operator): New predicate.
3784 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
3785 vector conditionals when the true and false values are constant
3786 vectors with all 0 bits or all 1 bits set.
3788 2017-02-06 Martin Sebor <msebor@redhat.com>
3790 PR tree-optimization/79376
3791 * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
3793 2017-02-06 Uros Bizjak <ubizjak@gmail.com>
3795 * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
3796 explicit subreg RTX with operand 1. Use VECTOR_MODE_P predicate
3797 to simplify split condition.
3799 2017-02-06 Jakub Jelinek <jakub@redhat.com>
3801 * omp-expand.c (oxpand_omp_atomic_fetch_op,
3802 expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
3805 2017-02-06 Segher Boessenkool <segher@kernel.crashing.org>
3807 PR rtl-optimization/68664
3808 * target.def (can_speculate_insn): New hook.
3809 * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
3810 * doc/tm.texi: Regenerate.
3811 * sched-rgn.c (can_schedule_ready_p): Use the new hook.
3812 * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
3813 (rs6000_sched_can_speculate_insn): New function.
3815 2017-02-06 Jakub Jelinek <jakub@redhat.com>
3817 PR tree-optimization/79284
3818 * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
3819 * tree-vect-stmts.c (vect_get_vec_def_for_operand,
3820 vectorizable_mask_load_store, vectorizable_operation,
3821 vect_is_simple_cond, get_same_sized_vectype): Use it instead
3822 of comparing TREE_CODE of a type against BOOLEAN_TYPE.
3823 * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
3824 vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
3825 * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
3826 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
3827 Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
3828 is_gimple_assign (stmt). Replace another such test with
3829 is_gimple_assign (stmt).
3831 2017-02-06 Georg-Johann Lay <avr@gjlay.de>
3834 * config/avr/avr.c (rtl-iter.h): Include it.
3835 (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
3836 (avr_legitimate_combined_insn): ...and implementation.
3838 2017-02-06 Dominik Vogt <vogt@linux.vnet.ibm.com>
3840 * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
3841 * config/s390/s390.c (s390_const_operand_ok)
3842 (s390_canonicalize_comparison, s390_extract_part)
3843 (s390_single_part, s390_contiguous_bitmask_nowrap_p)
3844 (s390_contiguous_bitmask_p, s390_rtx_costs)
3845 (legitimize_pic_address): Likewise.
3846 * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
3847 * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
3848 ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
3849 ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
3850 * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
3852 2017-02-06 Georg-Johann Lay <avr@gjlay.de>
3854 * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
3855 REGNO($0) == REGNO($1).
3857 2017-02-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3859 * config/s390/linux.h(SIZE_TYPE): Add comment.
3861 2017-02-06 Julian Brown <julian@codesourcery.com>
3862 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
3863 Virendra Pathak <virendra.pathak@broadcom.com>
3865 * config/aarch64/aarch64-cores.def: Change the scheduler
3867 * config/aarch64/aarch64.md: Include thunderx2t99.md.
3868 * config/aarch64/thunderx2t99.md: New file.
3870 2017-02-05 Gerald Pfeifer <gerald@pfeifer.com>
3872 * doc/standards.texi (Go Language): Update link to language
3875 2017-02-05 Jan Hubicka <hubicka@ucw.cz>
3877 * tree-eh.c (lower_resx): Sanitize profile.
3878 (cleanup_empty_eh_move_lp): Likewise.
3880 2017-02-05 Jan Hubicka <hubicka@ucw.cz>
3883 * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
3885 * cfgloopmanip.h (loop_version): Update prototype.
3886 * modulo-sched.c (sms_schedule): Update call of loop_version.
3887 * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
3888 * tree-parloops.c (gen_parallel_loop): Likewise.
3889 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
3890 * tree-ssa-loop-split.c (split_loop): Likewise.
3891 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
3892 * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
3894 2017-02-05 Martin Liska <mliska@suse.cz>
3897 * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
3899 (print_operand_address): Initialize a struct to zero.
3901 2017-02-05 Gerald Pfeifer <gerald@pfeifer.com>
3903 * doc/contrib.texi (Contributors): Refer to Hans Boehm's
3904 garbage collector only in textual form.
3906 2017-02-05 Gerald Pfeifer <gerald@pfeifer.com>
3908 * doc/extend.texi (x86 specific memory model extensions for
3909 transactional memory): Simplify a phrase.
3911 2017-02-05 Eric Botcazou <ebotcazou@adacore.com>
3914 * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
3915 'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
3916 (atomic_storedi_1): Likewise.
3918 2017-02-04 Jakub Jelinek <jakub@redhat.com>
3920 PR tree-optimization/79338
3921 * tree-parloops.c (gather_scalar_reductions): Don't call
3922 vect_analyze_loop_form for loop->inner before destroying loop's
3925 2017-02-03 Martin Sebor <msebor@redhat.com>
3927 PR tree-optimization/79327
3928 * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
3929 when precision has resulted in leading zeros.
3930 (format_integer): Adjust the likely counter to assume an unknown
3931 argument that may be zero is non-zero.
3933 2017-02-03 Jason Merrill <jason@redhat.com>
3936 * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
3937 avoid copying non-taken branch.
3939 2017-02-03 Jakub Jelinek <jakub@redhat.com>
3941 PR tree-optimization/79340
3942 * tree-vect-loop.c (vectorizable_reduction): Release
3943 vec_defs elements after safe_splicing them into other vectors.
3946 PR tree-optimization/79327
3947 * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
3948 true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
3950 (format_integer): Use wide_int_to_tree instead of build_int_cst
3951 + to_?hwi. If argmin is NULL, just set argmin and argmax to
3952 TYPE_{MIN,MAX}_VALUE of argtype. Simplify and fix computation
3953 of shortest and longest sequence.
3955 2017-02-03 Uros Bizjak <ubizjak@gmail.com>
3957 * config/i386/i386.c (dimode_scalar_chain::convert_reg):
3958 Use pextrd for TARGET_SSE4_1 when creating scalar copy.
3960 2017-02-03 Walter Lee <walt@tilera.com>
3963 * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
3964 after initial stackframe link reg save.
3965 * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
3967 2017-02-03 Jakub Jelinek <jakub@redhat.com>
3970 * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
3971 wu for stxssp alternative.
3973 2017-02-03 Martin Sebor <msebor@redhat.com>
3975 PR tree-optimization/79352
3976 * gimple-fold.c (get_range_strlen): Add argument.
3977 (get_range_strlen): Change return type to bool.
3978 (get_maxval_strlen): Pass in a dummy argument.
3979 * gimple-fold.h (get_range_strlen): Change return type to bool.
3980 * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
3981 * tree.h (array_at_struct_end_p): Add argument.
3982 * tree.c (array_at_struct_end_p): Handle it.
3984 2017-02-03 Martin Liska <mliska@suse.cz>
3987 * multiple_target.c (create_dispatcher_calls): Redirect edge
3988 from a caller of a dispatcher.
3989 (expand_target_clones): Make the clones local.
3990 (ipa_target_clone): Do both target clones and resolvers.
3991 (ipa_dispatcher_calls): Remove the pass.
3992 (pass_dispatcher_calls::gate): Likewise.
3993 (make_pass_dispatcher_calls): Likewise.
3994 * passes.def (pass_target_clone): Put as very first IPA early
3997 2017-02-03 Martin Liska <mliska@suse.cz>
3999 * symtab.c (symtab_node::binds_to_current_def_p): Bail out
4000 in case of a function with ifunc attribute.
4002 2017-02-03 Martin Liska <mliska@suse.cz>
4004 * cgraph.c (cgraph_node::dump): Dump function version info.
4005 * symtab.c (symtab_node::dump_base): Add missing new line.
4007 2017-02-02 Jan Hubicka <hubicka@ucw.cz>
4009 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
4010 (ifcombine_ifandif): Use it.
4012 2017-02-03 Martin Liska <mliska@suse.cz>
4014 * doc/invoke.texi: Document default value for
4015 use-after-scope-direct-emission-threshold.
4017 2017-02-03 Martin Liska <mliska@suse.cz>
4019 PR tree-optimization/79339
4020 * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
4021 (format_floating): Likewise.
4023 2017-02-03 Martin Liska <mliska@suse.cz>
4026 * ipa-prop.c (ipa_node_params_t::insert): Remove current
4028 (ipa_node_params_t::remove): Likewise.
4029 * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
4030 initialization from removed ipa_node_params_t::insert.
4031 (ipa_node_params::~ipa_node_params): Move from removed
4032 ipa_node_params_t::release.
4033 * symbol-summary.h (symbol_summary::m_released): New member.
4034 Do not release a summary twice. Do not allow to call finalizer
4035 for types of a summary that live in GGC memory.
4037 2017-02-02 Naveen H.S <Naveen.Hurugalawadi@cavium.com>
4039 * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
4042 2017-02-02 Martin Sebor <msebor@redhat.com>
4045 * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
4046 (format_string): Tighten up the range of output for non-constant
4047 strings and correct the expected range for wide non-constant strings.
4049 2017-02-02 Martin Sebor <msebor@redhat.com>
4051 * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
4054 * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
4056 (-fdump-tree-@var): Add to index and document how to come up
4057 with pass-specific option and dump file names.
4058 (-fdump-passes): Clarify where to look for output.
4060 2017-02-02 Jan Hubicka <hubicka@ucw.cz>
4063 * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
4064 statistics of the analyzed path; allow threading for speed when
4065 any of BBs along the path are optimized for speed.
4067 2017-02-02 Eric Botcazou <ebotcazou@adacore.com>
4070 * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
4071 settings of the virtual registers.
4074 2016-08-23 Dominik Vogt <vogt@linux.vnet.ibm.com>
4076 * explow.c (get_dynamic_stack_size): Take known alignment of stack
4077 pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
4080 2017-02-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4082 * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
4083 ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
4085 2017-02-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4087 * config/s390/s390.md: Add missing comments with the expanded
4089 * config/s390/vector.md: Likewise.
4090 * config/s390/vx-builtins.md: Likewise.
4092 2017-02-02 Jakub Jelinek <jakub@redhat.com>
4095 * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
4096 (fixuns_trunc<mode>di2): ... this, remove previous expander. Put all
4097 conditions on a single line.
4099 2017-02-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4101 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
4102 __S390_VX__ to __VX__.
4104 2017-02-01 Andrew Pinski <apinski@cavium.com>
4106 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
4107 stmt_info to record_stmt_cost.
4108 (vect_get_known_peeling_cost): Pass stmt_info if known to
4110 * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
4111 cpu_vector_cost field into
4112 scalar_int_stmt_cost and scalar_fp_stmt_cost. Split vec_stmt_cost
4113 field into vec_int_stmt_cost and vec_fp_stmt_cost.
4114 * config/aarch64/aarch64.c (generic_vector_cost): Update for the
4115 splitting of scalar_stmt_cost and vec_stmt_cost.
4116 (thunderx_vector_cost): Likewise.
4117 (cortexa57_vector_cost): LIkewise.
4118 (exynosm1_vector_cost): Likewise.
4119 (xgene1_vector_cost): Likewise.
4120 (thunderx2t99_vector_cost): Improve after the splitting of the two
4122 (aarch64_builtin_vectorization_cost): Update for the splitting of
4123 scalar_stmt_cost and vec_stmt_cost.
4125 2017-02-01 Torvald Riegel <triegel@redhat.com>
4126 Richard Henderson <rth@redhat.com>
4128 * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
4129 conditional on existance of a fast atomic load.
4130 * optabs-query.c (can_atomic_load_p): New function.
4131 * optabs-query.h (can_atomic_load_p): Declare it.
4132 * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
4133 no fast atomic load is available for the particular size of access.
4134 (expand_atomic_compare_and_swap): Likewise.
4135 (expand_atomic_load): Likewise.
4136 (expand_atomic_store): Likewise.
4137 (expand_atomic_fetch_op): Likewise.
4138 * testsuite/lib/target-supports.exp
4139 (check_effective_target_sync_int_128): Remove x86 because it provides
4140 no fast atomic load.
4141 (check_effective_target_sync_int_128_runtime): Likewise.
4143 2017-02-01 Richard Biener <rguenther@suse.de>
4145 * graphite.c: Include tree-vectorizer.h for find_loop_location.
4146 (graphite_transform_loops): Provide opt-info for optimized nests.
4147 * tree-parloop.c (parallelize_loops): Provide opt-info for
4150 2017-02-01 Richard Biener <rguenther@suse.de>
4153 * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
4156 2017-02-01 Richard Biener <rguenther@suse.de>
4158 PR tree-optimization/71824
4159 * graphite-scop-detection.c (scop_detection::build_scop_breadth):
4160 Verify the loops are valid in the merged SESE region.
4161 (scop_detection::can_represent_loop_1): Check analyzing the
4162 evolution of the number of iterations in the region succeeds.
4164 2017-01-31 Ian Lance Taylor <iant@golang.org>
4166 * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
4167 REG_ARGS_SIZE note to 32-bit push insns and call insn.
4169 2017-01-31 David Malcolm <dmalcolm@redhat.com>
4171 PR preprocessor/79210
4172 * input.c (get_substring_ranges_for_loc): Replace line_width
4173 assertion with error-handling.
4175 2017-01-31 Richard Biener <rguenther@suse.de>
4177 PR tree-optimization/77318
4178 * graphite-sese-to-poly.c (extract_affine): Fix assert.
4179 (create_pw_aff_from_tree): Take loop parameter.
4180 (add_condition_to_pbb): Pass loop of the condition to
4181 create_pw_aff_from_tree.
4183 2017-01-31 Jakub Jelinek <jakub@redhat.com>
4185 * config/s390/s390.c (s390_asan_shadow_offset): New function.
4186 (TARGET_ASAN_SHADOW_OFFSET): Redefine.
4188 2017-01-31 Michael Meissner <meissner@linux.vnet.ibm.com>
4192 * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
4194 (convert_int_to_float128): Likewise.
4195 * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
4196 (convert_int_to_float128): Likewise.
4197 * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
4198 (UNSPEC_IEEE128_CONVERT): Likewise.
4199 (floatsi<mode>2, FLOAT128 iterator): Bypass calling
4200 rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
4201 Use local variables for IBM extended format.
4202 (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
4203 (fix_trunc<mode>si2_fprs): Likewise.
4204 (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
4205 (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
4206 (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
4207 to know that we can now have integers of all sizes in vector
4209 (fix<uns>_<mode>di2_hw): Likewise.
4210 (float<uns>_<mode>si2_hw): Likewise.
4211 (fix_<mode>si2_hw): Likewise.
4212 (fixuns_<mode>si2_hw): Likewise.
4213 (float<uns>_<mode>di2_hw): Likewise.
4214 (float_<mode>di2_hw): Likewise.
4215 (float_<mode>si2_hw): Likewise.
4216 (floatuns_<mode>di2_hw): Likewise.
4217 (floatuns_<mode>si2_hw): Likewise.
4218 (xscvqp<su>wz_<mode>): Delete, no longer used.
4219 (xscvqp<su>dz_<mode>): Likewise.
4220 (xscv<su>dqp_<mode>): Likewise.
4221 (ieee128_mfvsrd_64bit): Likewise.
4222 (ieee128_mfvsrd_32bit): Likewise.
4223 (ieee128_mfvsrwz): Likewise.
4224 (ieee128_mtvsrw): Likewise.
4225 (ieee128_mtvsrd_64bit): Likewise.
4226 (ieee128_mtvsrd_32bit): Likewise.
4228 2017-01-31 Martin Liska <mliska@suse.cz>
4231 * ipa-prop.c (ipa_free_all_node_params): Call release method
4232 instead of ~sumbol_summary to not to trigger double times
4235 2017-01-31 Aldy Hernandez <aldyh@redhat.com>
4237 PR tree-optimization/71691
4238 * bitmap.h (class auto_bitmap): New.
4239 * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
4240 is_maybe_undefined instead of ssa_undefined_value_p.
4242 2017-01-31 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4244 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
4245 __S390_ARCH_LEVEL__ to __ARCH__.
4247 2017-01-31 Jakub Jelinek <jakub@redhat.com>
4249 PR tree-optimization/79267
4250 * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
4251 if should_remove_lhs_p is true.
4253 2017-01-30 Alexandre Oliva <aoliva@redhat.com>
4256 * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
4257 (add_alignment_attribute): New.
4258 (base_type_die): Add alignment attribute.
4259 (subrange_type_die): Likewise.
4260 (modified_type_die): Likewise.
4261 (gen_array_type_die): Likewise.
4262 (gen_descr_array_type_die: Likewise.
4263 (gen_enumeration_type_die): Likewise.
4264 (gen_subprogram_die): Likewise.
4265 (gen_variable_die): Likewise.
4266 (gen_field_die): Likewise.
4267 (gen_ptr_to_mbr_type_die): Likewise.
4268 (gen_struct_or_union_type_die): Likewise.
4269 (gen_subroutine_type_die): Likewise.
4270 (gen_typedef_die): Likewise.
4271 (base_type_cmp): Compare alignment attribute.
4273 2017-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4276 * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
4277 (setb_unsigned) New pattern for setb with CCUNS.
4278 * config/rs6000/rs6000.c (expand_block_compare): Use a different
4279 subfc./subfe sequence to avoid overflow problems. Generate a
4280 shorter sequence with cmpld/setb for power9.
4281 * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
4282 for generating subfc. instruction.
4283 (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
4284 now uses this instruction.
4286 2017-01-30 Ian Lance Taylor <iant@google.com>
4289 * dwarf2out.c (gen_type_die_with_usage): When picking a variant
4290 for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
4292 2017-01-30 Martin Sebor <msebor@redhat.com>
4294 * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
4295 Move constant to the right of a relational operator.
4296 (get_mpfr_format_length, format_character, format_string): Ditto.
4297 (should_warn_p, maybe_warn): Same.
4299 * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
4301 2017-01-30 Maxim Ostapenko <m.ostapenko@samsung.com>
4304 * asan.c (get_translation_unit_decl): Remove function.
4305 (asan_add_global): Force has_dynamic_init to zero in LTO mode.
4307 2017-01-30 Martin Liska <mliska@suse.cz>
4309 PR gcov-profile/79259
4310 * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
4313 2017-01-30 Martin Liska <mliska@suse.cz>
4316 * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
4317 Initialize variables with NULL value.
4319 2017-01-30 Richard Earnshaw <rearnsh@arm.com>
4322 * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
4324 * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
4326 2017-01-30 Richard Biener <rguenther@suse.de>
4328 PR tree-optimization/79276
4329 * tree-vrp.c (process_assert_insertions): Properly adjust common
4330 when removing a duplicate.
4332 * gcc.dg/torture/pr79276.c: New testcase.
4334 2017-01-30 Richard Biener <rguenther@suse.de>
4336 PR tree-optimization/79256
4337 * targhooks.c (default_builtin_vector_alignment_reachable): Honor
4338 BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
4340 * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
4342 2017-01-30 Dominik Vogt <vogt@linux.vnet.ibm.com>
4345 * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
4346 ("*r<noxa>sbg_<mode>_sll_bitmask")
4347 ("*extzv_<mode>_srl<clobbercc_or_nocc>")
4348 ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
4349 Use contiguous_bitmask_nowrap_operand.
4351 2017-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4354 * config/rs6000/altivec.h (vec_xl): Revise #define.
4355 (vec_xst): Likewise.
4357 2017-01-27 Uros Bizjak <ubizjak@gmail.com>
4359 * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
4361 2017-01-27 Bernd Schmidt <bschmidt@redhat.com>
4363 PR rtl-optimization/79194
4364 * cprop.c (one_cprop_pass): Move deletion of code after unconditional
4365 traps before call to bypass_conditional_jumps.
4367 2017-01-27 Vladimir Makarov <vmakarov@redhat.com>
4369 PR tree-optimization/71374
4370 * lra-constraints.c (check_conflict_input_operands): New.
4371 (match_reload): Use it.
4373 2017-01-27 Vladimir Makarov <vmakarov@redhat.com>
4376 * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
4377 account to calculate conflict_set.
4379 2017-01-27 Bin Cheng <bin.cheng@arm.com>
4381 PR rtl-optimization/78559
4382 * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
4383 other_insn in combine.
4385 2017-01-27 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4387 * builtin-types.def: Use unsigned_char_type_node for BT_UINT8. Use
4388 uint16_type_node for BT_UINT16.
4390 2017-01-27 David Malcolm <dmalcolm@redhat.com>
4392 * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
4393 "RTL Tests" to menu.
4394 (GIMPLE Tests): New node.
4395 (RTL Tests): New node.
4397 2017-01-27 Richard Biener <rguenther@suse.de>
4399 PR tree-optimization/79245
4400 * tree-loop-distribution.c (distribute_loop): Apply cost
4401 modeling also to detected patterns.
4403 2017-01-27 Richard Biener <rguenther@suse.de>
4405 PR tree-optimization/71433
4406 * tree-vrp.c (register_new_assert_for): Revert earlier changes.
4407 (compare_assert_loc): New function.
4408 (process_assert_insertions): Sort and optimize assert locations
4409 to remove duplicates and push down identical assertions on
4410 edges to their destination block.
4412 2017-01-27 Richard Biener <rguenther@suse.de>
4414 PR tree-optimization/79244
4415 * tree-vrp.c (remove_range_assertions): Forcefully propagate
4416 out SSA names even if abnormal.
4418 2017-01-27 Jakub Jelinek <jakub@redhat.com>
4420 * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
4421 * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
4422 instead of MPFR_RNDN.
4424 2017-01-27 Richard Earnshaw <rearnsha@arm.com>
4427 * arm.c (arm_option_override): Don't call build_target_option_node
4428 until after doing all option overrides.
4429 (arm_valid_target_attribute_tree): Likewise.
4431 2017-01-27 Martin Liska <mliska@suse.cz>
4433 * doc/invoke.texi (-fprofile-arcs): Document profiling support
4434 for {cd}tors and C++ {cd}tors.
4436 2017-01-27 Dominik Vogt <vogt@linux.vnet.ibm.com>
4438 * config/s390/s390.md ("*setmem_long_and")
4439 ("*setmem_long_and_31z"): Use zero_extend instead of and.
4441 2017-01-26 Martin Sebor <msebor@redhat.com>
4443 * gimple-ssa-sprintf.c (format_floating): Simplify the computation
4446 2017-01-26 Martin Sebor <msebor@redhat.com>
4448 * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
4449 HAVE_DFmode before using XFmode or DFmode.
4450 (parse_directive): Avoid using the z length modifier to avoid
4451 the ISO C++98 does not support the ‘z’ gnu_printf length modifier.
4454 * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
4455 to accept adjustment as an array.
4456 (get_int_range): New function.
4457 (struct directive): Make width and prec arrays.
4458 (directive::set_width, directive::set_precision): Call get_int_range.
4459 (format_integer, format_floating): Handle width and precision ranges.
4460 (format_string, parse_directive): Same.
4462 2017-01-26 Jakub Jelinek <jakub@redhat.com>
4465 * dwarf2out.c (generate_skeleton_bottom_up): For children with
4466 comdat_type_p set, just clone them, but keep the children in the
4470 * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
4471 which have direct callers with -fvar-tracking-assignments enabled
4473 (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
4474 inside of type units.
4476 2017-01-26 Martin Sebor <msebor@redhat.com>
4479 * gimple-ssa-sprintf.c (struct result_range): Add likely and
4481 (struct format_result): Replace number_chars, number_chars_min,
4482 and number_chars_max with a single member of struct result_range.
4484 (format_result::operator+=): Adjust.
4485 (struct fmtresult): Remove bounded. Handle likely and unlikely
4487 (fmtresult::adjust_for_width_or_precision): New function.
4488 (fmtresult:type_max_digits): New function.
4489 (bytes_remaining): Handle likely and unlikely counters.
4490 (min_bytes_remaining): Remove.
4491 (format_percent): Simplify.
4492 (format_integer, format_floating): Set likely and unlikely counters.
4493 (get_string_length, format_character, format_string): Same.
4494 (format_plain, should_warn_p): New function.
4495 (maybe_warn): Call should_warn_p. Update diagnostic messages
4496 and handle those for all directives, including plain strings.
4497 (format_directive): Handle likely and unlikely counters.
4498 Remove unnecessary quoting from diagnostics. Add an informational
4500 (add_bytes): Remove.
4501 (pass_sprintf_length::compute_format_length): Simplify.
4502 (try_substitute_return_value): Handle likely and unlikely counters.
4504 2017-01-26 Carl Love <cel@us.ibm.com>
4506 * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
4507 bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
4509 2017-01-26 Vladimir Makarov <vmakarov@redhat.com>
4512 * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
4513 endianess for subregs into account.
4514 * lra-constraints.c (lra_constraints): Do risky transformations
4515 always on the first iteration.
4516 * lra-lives.c (check_pseudos_live_through_calls): Add arg
4517 last_call_used_reg_set.
4518 (process_bb_lives): Define and use last_call_used_reg_set.
4519 * lra.c (lra): Always continue after lra_constraints on the first
4522 2017-01-26 Kirill Yukhin <kirill.yukhin@gmail.com>
4524 * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
4526 * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
4528 2017-01-26 Jakub Jelinek <jakub@redhat.com>
4530 * config/i386/avx512fintrin.h (_ktest_mask16_u8,
4531 _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
4532 * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
4533 _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
4534 * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
4535 __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
4536 OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
4537 * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
4538 (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
4540 2017-01-26 Marek Polacek <polacek@redhat.com>
4543 * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
4544 for the third operand.
4546 2017-01-26 Jakub Jelinek <jakub@redhat.com>
4549 * omp-low.c (struct omp_context): Add simt_stmt field.
4550 (scan_omp_for): Return omp_context *.
4551 (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
4552 context to the _simt_ SIMD stmt.
4553 (lower_omp_for): For combined SIMD with sibling _simt_
4554 SIMD, make sure to use the same decls in _looptemp_
4555 clauses as in the sibling.
4557 2017-01-26 David Sherwood <david.sherwood@arm.com>
4560 * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
4563 2017-01-26 Jakub Jelinek <jakub@redhat.com>
4566 * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
4567 emit fld b; fld a; if possible.
4569 * brig-builtins.def: Update copyright years.
4570 * config/arm/arm_acle_builtins.def: Update copyright years.
4572 2017-01-25 Michael Meissner <meissner@linux.vnet.ibm.com>
4575 * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
4576 constraint instead of o for the stxsd instruction.
4578 2017-01-25 Carl Love <cel@us.ibm.com>
4580 * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
4581 of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
4583 2017-01-25 Jonathan Wakely <jwakely@redhat.com>
4585 * doc/invoke.texi (C++ Dialect Options): Fix typo.
4587 2017-01-25 Richard Biener <rguenther@suse.de>
4589 PR tree-optimization/69264
4590 * target.def (vector_alignment_reachable): Improve documentation.
4591 * doc/tm.texi: Regenerate.
4592 * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
4594 * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
4595 earlier changes with respect to TYPE_USER_ALIGN.
4596 (vector_alignment_reachable_p): Likewise. Improve dumping.
4598 2016-01-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4601 * config/arm/arm.md (xordi3): Force constant operand into a register
4604 2016-01-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4606 * doc/invoke.texi (-fstore-merging): Correct default optimization
4607 levels at which it is enabled.
4608 (-O): Move -fstore-merging from list to...
4611 2017-01-25 Richard Biener <rguenther@suse.de>
4614 * omp-expand.c: Include debug.h.
4615 (expand_omp_taskreg): Make sure to generate early debug before
4616 outlining anything from a function.
4617 (expand_omp_target): Likewise.
4618 (grid_expand_target_grid_body): Likewise.
4620 2017-01-25 Maxim Ostapenko <m.ostapenko@samsung.com>
4623 * asan.c (get_translation_unit_decl): New function.
4624 (asan_add_global): Extract modules file name from globals
4625 TRANSLATION_UNIT_DECL name.
4627 2017-01-24 Eric Botcazou <ebotcazou@adacore.com>
4630 * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
4631 for long calls with APCS frame and VFP.
4633 2017-01-24 David Malcolm <dmalcolm@redhat.com>
4635 * cfg.c (original_copy_tables_initialized_p): New function.
4636 * cfg.h (original_copy_tables_initialized_p): New decl.
4637 * cfgrtl.c (relink_block_chain): Guard the call to
4638 free_original_copy_tables with a call to
4639 original_copy_tables_initialized_p.
4640 * cgraph.h (symtab_node::native_rtl_p): New decl.
4641 * cgraphunit.c (symtab_node::native_rtl_p): New function.
4642 (symtab_node::needed_p): Don't assert for early assembly output
4643 for __RTL functions.
4644 (cgraph_node::finalize_function): Set "force_output" for __RTL
4646 (cgraph_node::analyze): Bail out early for __RTL functions.
4647 (analyze_functions): Update assertion to support __RTL functions.
4648 (cgraph_node::expand): Bail out early for __RTL functions.
4649 * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
4651 * function.h (struct function): Update comment for field
4653 * gimple-expr.c: Include "tree-pass.h".
4654 (gimple_has_body_p): Return false for __RTL functions.
4655 * Makefile.in (OBJS): Add run-rtl-passes.o.
4656 * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
4658 (gcc::pass_manager::get_clean_slate): New accessor.
4659 * passes.c: Include "insn-addr.h".
4660 (should_skip_pass_p): Add logging. Update logic for running
4661 "expand" to be compatible with both __GIMPLE and __RTL. Guard
4662 property-provider override so it is only done for gimple passes.
4664 (skip_pass): New function.
4665 (execute_one_pass): Call skip_pass when skipping passes.
4666 * read-md.c (md_reader::read_char): Support filtering
4667 the input to a subset of line numbers.
4668 (md_reader::md_reader): Initialize fields
4669 m_first_line and m_last_line.
4670 (md_reader::read_file_fragment): New function.
4671 * read-md.h (md_reader::read_file_fragment): New decl.
4672 (md_reader::m_first_line): New field.
4673 (md_reader::m_last_line): New field.
4674 * read-rtl-function.c (function_reader::create_function): Only
4675 create cfun if it doesn't already exist. Set PROP_rtl on cfun's
4676 curr_properties. Set DECL_INITIAL to a dummy block.
4677 (read_rtl_function_body_from_file_range): New function.
4678 * read-rtl-function.h (read_rtl_function_body_from_file_range):
4680 * run-rtl-passes.c: New file.
4681 * run-rtl-passes.h: New file.
4683 2017-01-24 Jeff Law <law@redhat.com>
4685 * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
4688 2017-01-24 Bin Cheng <bin.cheng@arm.com>
4690 PR tree-optimization/79159
4691 * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
4692 (record_nonwrapping_iv): Improve boundary using above function if no
4693 value range information.
4695 2017-01-24 Pekka Jääskeläinen <pekka@parmance.com>
4696 Martin Jambor <mjambor@suse.cz>
4698 * brig-builtins.def: New file.
4699 * builtins.def (DEF_HSAIL_BUILTIN): New macro.
4700 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
4701 (DEF_HSAIL_SAT_BUILTIN): Likewise.
4702 (DEF_HSAIL_INTR_BUILTIN): Likewise.
4703 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
4704 * builtin-types.def (BT_INT8): New.
4705 (BT_INT16): Likewise.
4706 (BT_UINT8): Likewise.
4707 (BT_UINT16): Likewise.
4708 (BT_FN_ULONG): Likewise.
4709 (BT_FN_UINT_INT): Likewise.
4710 (BT_FN_UINT_ULONG): Likewise.
4711 (BT_FN_UINT_LONG): Likewise.
4712 (BT_FN_UINT_PTR): Likewise.
4713 (BT_FN_ULONG_PTR): Likewise.
4714 (BT_FN_INT8_FLOAT): Likewise.
4715 (BT_FN_INT16_FLOAT): Likewise.
4716 (BT_FN_UINT32_FLOAT): Likewise.
4717 (BT_FN_UINT16_FLOAT): Likewise.
4718 (BT_FN_UINT8_FLOAT): Likewise.
4719 (BT_FN_UINT64_FLOAT): Likewise.
4720 (BT_FN_UINT16_UINT32): Likewise.
4721 (BT_FN_UINT32_UINT16): Likewise.
4722 (BT_FN_UINT16_UINT16_UINT16): Likewise.
4723 (BT_FN_INT_PTR_INT): Likewise.
4724 (BT_FN_UINT_PTR_UINT): Likewise.
4725 (BT_FN_LONG_PTR_LONG): Likewise.
4726 (BT_FN_ULONG_PTR_ULONG): Likewise.
4727 (BT_FN_VOID_UINT64_UINT64): Likewise.
4728 (BT_FN_UINT8_UINT8_UINT8): Likewise.
4729 (BT_FN_INT8_INT8_INT8): Likewise.
4730 (BT_FN_INT16_INT16_INT16): Likewise.
4731 (BT_FN_INT_INT_INT): Likewise.
4732 (BT_FN_UINT_FLOAT_UINT): Likewise.
4733 (BT_FN_FLOAT_UINT_UINT): Likewise.
4734 (BT_FN_ULONG_UINT_UINT): Likewise.
4735 (BT_FN_ULONG_UINT_PTR): Likewise.
4736 (BT_FN_ULONG_ULONG_ULONG): Likewise.
4737 (BT_FN_UINT_UINT_UINT): Likewise.
4738 (BT_FN_VOID_UINT_PTR): Likewise.
4739 (BT_FN_UINT_UINT_PTR: Likewise.
4740 (BT_FN_UINT32_UINT64_PTR): Likewise.
4741 (BT_FN_INT_INT_UINT_UINT): Likewise.
4742 (BT_FN_UINT_UINT_UINT_UINT): Likewise.
4743 (BT_FN_UINT_UINT_UINT_PTR): Likewise.
4744 (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
4745 (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
4746 (BT_FN_LONG_LONG_UINT_UINT): Likewise.
4747 (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
4748 (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
4749 (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
4750 (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
4751 (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
4752 (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
4753 * doc/frontends.texi: List BRIG FE.
4754 * doc/install.texi (Testing): Add BRIG tesring requirements.
4755 * doc/invoke.texi (Overall Options): Mention BRIG.
4756 * doc/standards.texi (Standards): Doucment BRIG HSA version.
4758 2017-01-24 Richard Biener <rguenther@suse.de>
4760 PR translation/79208
4761 * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
4763 2017-01-24 Martin Jambor <mjambor@suse.cz>
4766 * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
4767 * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
4770 2017-01-24 Aldy Hernandez <aldyh@redhat.com>
4773 * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
4774 casts from signed to unsigned really don't have a range.
4776 2017-01-24 Markus Trippelsdorf <markus@trippelsdorf.de>
4778 * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
4779 GMP_RNDx for compatiblity.
4781 2017-01-24 Martin Liska <mliska@suse.cz>
4784 * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
4785 that would prevent us to call alloca with -1 as argument.
4787 2017-01-24 Jakub Jelinek <jakub@redhat.com>
4789 * dwarf2out.c (output_compilation_unit_header, output_file_names):
4790 Avoid -Wformat-security warning.
4792 2017-01-23 Andrew Pinski <apinski@cavium.com>
4794 * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
4797 2017-01-23 Martin Sebor <msebor@redhat.com>
4800 * gimple-ssa-sprintf.c (warn_level): New global.
4801 (format_integer): Use it here and throughout the rest of the file.
4802 Use the same switch to compute sign as base.
4803 (maybe_warn): New function.
4804 (format_directive): Factor out warnings into maybe_warn.
4805 Add debugging output. Use warn_level.
4806 (add_bytes): Use warn_level.
4807 (pass_sprintf_length::compute_format_length): Add debugging output.
4808 (try_substitute_return_value): Same.
4809 (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
4812 * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
4813 (struct fmtresult, format_integer, format_floating): Adjust.
4814 (fmtresult::fmtresult): Set max correctly in two argument ctor.
4815 (get_string_length, format_string,format_directive): Same.
4816 (pass_sprintf_length::compute_format_length): Same.
4817 (try_substitute_return_value): Simplify slightly.
4820 * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
4821 (fmtresult::operator+=): Outlined.
4822 (struct fmtresult): Add ctors.
4823 (struct conversion_spec): Rename...
4824 (struct directive): ...to this. Add and remove data members.
4825 (directive::set_width, directive::set_precision): New functions.
4826 (format_percent): Use fmtresult ctor.
4827 (get_width_and_precision): Remove.
4828 (format_integer): Make naming changes. Avoid computing width and
4830 (format_floating): Same. Adjust indentation.
4831 (format_character, format_none): New functions.
4832 (format_string): Moved character handling to format_character.
4833 (format_directive): Remove arguments, change return type.
4834 (parse_directive): New function.
4835 (pass_sprintf_length::compute_format_length): Move directive
4836 parsing to parse_directive.
4838 2017-01-23 Jakub Jelinek <jakub@redhat.com>
4840 * tree.h (assign_assembler_name_if_neeeded): Rename to ...
4841 (assign_assembler_name_if_needed): ... this.
4842 * tree.c (assign_assembler_name_if_neeeded): Rename to ...
4843 (assign_assembler_name_if_needed): ... this.
4844 (free_lang_data_in_cgraph): Adjust callers.
4845 * cgraphunit.c (cgraph_node::analyze): Likewise.
4846 * omp-expand.c (expand_omp_taskreg, expand_omp_target):
4849 2017-01-23 Richard Biener <rguenther@suse.de>
4851 PR tree-optimization/79088
4852 PR tree-optimization/79188
4853 * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
4854 resetting loop bounds after last path deletion. Reset loop
4855 bounds of the target loop, make code match the comments.
4856 * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
4857 Make sure loops need no fixups.
4859 2017-01-23 Kelvin Nilsen <kelvin@gcc.gnu.org>
4861 * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
4862 exponent support with double type for first argument.
4863 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
4864 type returned by __builtin_vec_extract_sig,
4865 __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
4866 functions from "vector int" to "vector unsigned int" or from
4867 "vector long long int" to "vector unsigned long long int".
4868 Changed type returned by __builtin_vec_extract_exp,
4869 __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
4870 functions from "vector int" to "vector unsigned int" or from
4871 "vector long long int" to "vector unsigned long long int".
4872 Changed return type of __builtin_vec_test_data_class,
4873 __builtin_vec_test_data_class_sp, and
4874 __builtin_vec_test_data_class_dp from "vector int" to
4875 "vector bool int" or from "vector long long int" to "vector bool
4876 long long int" and changed second argument type from "unsigned
4877 int" to "int". Added new overloaded function forms "vector float
4878 __builtin_vec_insert_exp (vector float, vector unsigned int)" and
4879 "vector float __builtin_vec_insert_exp_sp (vector float, vector
4880 unsigned int)" and "vector double __builtin_vec_insert_exp (vector
4881 double, vector unsigned long long int)" and "vector double
4882 __builtin_vec_insert_exp_dp (vector double, vector unsigned long
4883 long int)". Changed return type of
4884 __builtin_scalar_test_data_class and
4885 __builtin_scalar_test_data_class_sp and
4886 __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
4887 int" and changed second argument from "unsigned int" to "int".
4888 Changed type returned by __builtin_scalar_test_neg,
4889 __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
4890 from "int" to "bool int". Added new overloaded function form
4891 "double __builtin_scalar_insert_exp (double, unsigned long long int)".
4892 * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
4893 exponent double-precision with floating point first argument.
4894 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4895 documentation of scalar_test_data_class, scalar_test_neg,
4896 scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
4897 vector_extract_exp, vec_extract_sig, vec_insert_exp, and
4898 vec_test_data_class built-in functions to reflect refinements in
4899 their type signatures.
4901 2017-01-23 Andreas Tobler <andreast@gcc.gnu.org>
4903 * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
4905 (aarch64_elf_asm_destructor): Likewise.
4907 2017-01-23 Bernd Schmidt <bschmidt@redhat.com>
4909 PR rtl-optimization/78634
4910 * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
4911 (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
4912 * ifcvt.c (noce_try_cmove): Add missing cost check.
4914 PR rtl-optimization/71724
4915 * combine.c (if_then_else_cond): Look for situations where it is
4916 beneficial to undo the work of one of the recursive calls.
4918 2017-01-23 Bin Cheng <bin.cheng@arm.com>
4920 PR tree-optimization/70754
4921 * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
4922 (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE. Insert
4923 combined stmt before it if not NULL.
4924 (combine_chains): Process refs reversely and compute dominance point
4927 2017-01-23 Martin Liska <mliska@suse.cz>
4929 PR tree-optimization/79196
4930 * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
4931 (fold_strstr_to_strncmp): ... this. Fold the pattern to strncmp
4933 (strlen_optimize_stmt): Call the renamed function.
4935 2017-01-23 Michael Matz <matz@suse.de>
4937 PR tree-optimization/78384
4938 * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
4940 2017-01-23 Richard Biener <rguenther@suse.de>
4942 PR tree-optimization/79186
4943 * tree-vrp.c (register_new_assert_for): Make sure we've seen
4944 both incoming edges before moving an assert.
4946 2017-01-23 Martin Jambor <mjambor@suse.cz>
4948 * ipa-prop.c (load_from_param_1): Removed.
4949 (load_from_unmodified_param): Bits from load_from_param_1 put back
4951 (load_from_param): Removed.
4952 (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
4953 with stmt. Reverted back to use of load_from_unmodified_param.
4955 2017-01-23 Martin Jambor <mjambor@suse.cz>
4958 * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
4959 (ipa_node_params): Annotate with GTY((for_user)). Make descriptors
4960 field a pointer to garbage collected vector, mark lattices and
4961 ipcp_orig_node with GTY((skip)).
4962 (ipa_get_param_count): Adjust to descriptors being a pointer.
4963 (ipa_get_param): Likewise.
4964 (ipa_get_type): Likewise.
4965 (ipa_get_param_move_cost): Likewise.
4966 (ipa_set_param_used): Likewise.
4967 (ipa_get_controlled_uses): Likewise.
4968 (ipa_set_controlled_uses): Likewise.
4969 (ipa_is_param_used): Likewise.
4970 (ipa_node_params_t): Move into garbage collector. New methods insert
4972 (ipa_node_params_sum): Annotate wth GTY(()).
4973 (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
4975 (ipa_load_from_parm_agg): Adjust declaration.
4976 * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
4977 * ipa-profile.c (ipa_profile): Likewise.
4978 * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
4979 (ipa_populate_param_decls): Make descriptors parameter garbage
4981 (ipa_dump_param): Adjust to descriptors being a pointer.
4982 (ipa_alloc_node_params): Likewise.
4983 (ipa_initialize_node_params): Likewise.
4984 (load_from_param_1): Make descriptors parameter garbage collected.
4985 (load_from_unmodified_param): Likewise.
4986 (load_from_param): Likewise.
4987 (ipa_load_from_parm_agg): Likewise.
4988 (ipa_node_params::~ipa_node_params): Removed.
4989 (ipa_free_all_node_params): Remove call to delete operator.
4990 (ipa_node_params_t::insert): New.
4991 (ipa_node_params_t::remove): Likewise.
4992 (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
4993 copy known_csts and known_contexts vectors.
4994 (ipa_read_node_info): Adjust to descriptors being a pointer.
4995 (ipcp_modif_dom_walker): Make m_descriptors field garbage
4997 (ipcp_transform_function): Make descriptors variable garbage
5000 2017-01-23 Andrew Senkevich <andrew.senkevich@intel.com>
5002 * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
5003 * config/i386/avx512dqintrin.h: Ditto.
5004 * config/i386/avx512fintrin.h: Ditto.
5005 * gcc/config/i386/i386.c: Handle new builtins.
5006 * config/i386/i386-builtin.def: Add new builtins.
5007 * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
5008 (UNSPEC_KORTEST, UNSPEC_KTEST): New.
5010 2017-01-23 Jakub Jelinek <jakub@redhat.com>
5011 Martin Liska <mliska@suse.cz>
5013 * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
5014 * asan.c (asan_expand_poison_ifn): Support stores and use
5015 appropriate ASAN report function.
5016 * internal-fn.c (expand_ASAN_POISON_USE): New function.
5017 * internal-fn.def (ASAN_POISON_USE): Declare.
5018 * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
5019 (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
5020 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
5021 ASAN_POISON calls w/o LHS.
5022 * tree-ssa.c (execute_update_addresses_taken): Create clobber
5023 for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
5024 from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
5025 * gimplify.c (asan_poison_variables): Add attribute
5026 use_after_scope_memory to variables that really needs to live
5028 * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
5029 having the attribute.
5031 2017-01-23 Martin Liska <mliska@suse.cz>
5033 * asan.c (create_asan_shadow_var): New function.
5034 (asan_expand_poison_ifn): Likewise.
5035 * asan.h (asan_expand_poison_ifn): New declaration.
5036 * internal-fn.c (expand_ASAN_POISON): Likewise.
5037 * internal-fn.def (ASAN_POISON): New builtin.
5038 * sanopt.c (pass_sanopt::execute): Expand
5039 asan_expand_poison_ifn.
5040 * tree-inline.c (copy_decl_for_dup_finish): Make function
5042 * tree-inline.h (copy_decl_for_dup_finish): Likewise.
5043 * tree-ssa.c (is_asan_mark_p): New function.
5044 (execute_update_addresses_taken): Rewrite local variables
5045 (identified just by use-after-scope as addressable) into SSA.
5047 2017-01-22 Gerald Pfeifer <gerald@pfeifer.com>
5049 * doc/install.texi (Specific): opensource.apple.com uses https
5050 now. Remove trailing slash.
5052 2017-01-22 Gerald Pfeifer <gerald@pfeifer.com>
5054 * README.Portability: Remove note on an Irix compatibility issue.
5056 2017-01-22 Dimitry Andric <dim@FreeBSD.org>
5058 * gcov.c (INCLUDE_ALGORITHM): Define.
5059 (INCLUDE_VECTOR): Define.
5060 No longer include <vector> and <algorithm> directly.
5062 2017-01-21 Gerald Pfeifer <gerald@pfeifer.com>
5064 * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
5066 * doc/invoke.texi (Code Gen Options): Ditto.
5068 2017-01-21 Jan Hubicka <hubicka@ucw.cz>
5071 * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
5073 2017-01-21 Bernd Schmidt <bschmidt@redhat.com>
5075 rtl-optimization/79125
5076 * cprop.c (local_cprop_pass): Handle cases where we make an
5079 2017-01-20 Segher Boessenkool <segher@kernel.crashing.org>
5083 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
5084 read from, for big endian.
5086 2017-01-20 Jiong Wang <jiong.wang@arm.com>
5088 * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
5089 register pauth builtins for LP64 only.
5091 2017-01-20 Marek Polacek <polacek@redhat.com>
5094 * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
5097 2017-01-20 Alexander Monakov <amonakov@ispras.ru>
5099 * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
5101 * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
5102 * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
5103 * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
5105 2017-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5108 * config/arm/arm.c (neon_valid_immediate): Reject vector constants
5109 in big-endian mode when they are not a single duplicated value.
5111 2017-01-20 Richard Biener <rguenther@suse.de>
5113 * BASE-VER: Bump to 7.0.1.
5115 2017-01-20 Alexander Monakov <amonakov@ispras.ru>
5117 * omp-low.c (omplow_simd_context): New struct. Use it...
5118 (lower_rec_simd_input_clauses): ...here and...
5119 (lower_rec_input_clauses): ...here to hold common data. Adjust all
5120 references to idx, lane, max_vf, is_simt.
5122 2017-01-20 Graham Markall <graham.markall@embecosm.com>
5124 * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
5127 2017-01-20 Martin Jambor <mjambor@suse.cz>
5129 * hsa.h: Renaed to hsa-common.h. Adjusted a comment.
5130 * hsa.c: Renaed to hsa-common.c. Change include of gt-hsa.h to
5132 * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
5133 (GTFILES): Rename hsa.c to hsa-common.c.
5134 * hsa-brig.c: Change include of hsa.h to hsa-common.h.
5135 * hsa-dump.c: Likewise.
5136 * hsa-gen.c: Likewise.
5137 * hsa-regalloc.c: Likewise.
5138 * ipa-hsa.c: Likewise.
5139 * omp-expand.c: Likewise.
5140 * omp-low.c: Likewise.
5141 * toplev.c: Likewise.
5143 2017-01-20 Marek Polacek <polacek@redhat.com>
5146 * doc/invoke.texi: Document -Wduplicated-branches.
5147 * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
5148 COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
5149 POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
5150 STATEMENT_LIST, and RETURN_EXPR. For non-pure non-const functions
5151 return 0 only when not OEP_LEXICOGRAPHIC.
5152 (fold_build_cleanup_point_expr): Use the expression
5153 location when building CLEANUP_POINT_EXPR.
5154 * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
5155 * tree.c (add_expr): Handle error_mark_node.
5157 2017-01-20 Martin Liska <mliska@suse.cz>
5160 * tree-profile.c (init_ic_make_global_vars): Do not call
5162 (gimple_init_gcov_profiler): Likewise.
5164 2017-01-20 Martin Liska <mliska@suse.cz>
5167 * cgraph.h (maybe_create_reference): Remove argument and
5169 * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
5171 * ipa-cp.c (create_specialized_node): Likewise.
5172 * symtab.c (symtab_node::maybe_create_reference): Handle
5173 VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
5175 2017-01-20 Martin Liska <mliska@suse.cz>
5177 * read-rtl-function.c (function_reader::create_function): Use
5178 build_decl instread of build_decl_stat.
5180 2017-01-20 Andrew Senkevich <andrew.senkevich@intel.com>
5182 * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
5183 * config/i386/avx512dqintrin.h: Ditto.
5184 * config/i386/avx512fintrin.h: Ditto.
5185 * config/i386/i386-builtin-types.def: Add new types.
5186 * gcc/config/i386/i386.c: Handle new types.
5187 * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
5188 (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
5189 (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
5190 (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
5191 (__builtin_ia32_kshiftridi): New.
5192 * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
5194 2017-01-19 Segher Boessenkool <segher@kernel.crashing.org>
5198 * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
5199 define to rs6000_init_stack_protect_guard.
5200 (rs6000_init_stack_protect_guard): New function.
5202 2017-01-19 Matthew Fortune <matthew.fortune@imgtec.com>
5203 Yunqiang Su <yunqiang.su@imgtec.com>
5205 * config.gcc (supported_defaults): Add madd4.
5206 (with_madd4): Add validation.
5207 (all_defaults): Add madd4.
5208 * config/mips/mips.opt (mmadd4): New option.
5209 * gcc/config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
5211 (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
5213 (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
5214 (ISA_HAS_FUSED_MADD4): Likewise.
5215 * gcc/doc/invoke.texi (-mmadd4): Document the new option.
5216 * gcc/doc/install.texi (--with-madd4): Document the new option.
5218 2017-01-19 Jiong Wang <jiong.wang@arm.com>
5220 * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
5221 entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
5222 AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
5223 (aarch64_init_pauth_hint_builtins): New.
5224 (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
5225 (aarch64_expand_builtin): Expand new builtins.
5227 2017-01-19 Jiong Wang <jiong.wang@arm.com>
5229 * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
5230 * combine-stack-adj.c (no_unhandled_cfa): Handle
5231 REG_CFA_TOGGLE_RA_MANGLE.
5232 * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
5233 * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
5234 info for return address signing.
5235 (aarch64_expand_epilogue): Likewise.
5237 2017-01-19 Jiong Wang <jiong.wang@arm.com>
5239 * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
5240 * config/aarch64/aarch64-protos.h
5241 (aarch64_return_address_signing_enabled): New declaration.
5242 * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
5244 (aarch64_expand_prologue): Sign return address before it's pushed onto
5246 (aarch64_expand_epilogue): Authenticate return address fetched from
5248 (aarch64_override_options): Sanity check for ILP32 and ISA level.
5249 (aarch64_attributes): New function attributes for "sign-return-address".
5250 * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
5251 UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
5252 ("*do_return"): Generate combined instructions according to key index.
5253 ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
5254 * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
5256 (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
5257 * config/aarch64/aarch64.opt (msign-return-address=): New.
5258 * doc/extend.texi (AArch64 Function Attributes): Documents
5259 "sign-return-address=".
5260 * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
5262 2017-01-19 Matthew Fortune <matthew.fortune@imgtec.com>
5264 * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
5265 overall option summary.
5267 2017-01-19 Jiong Wang <jiong.wang@arm.com>
5269 * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
5270 * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
5271 AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
5272 * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
5274 2017-01-19 Michael Meissner <meissner@linux.vnet.ibm.com>
5276 * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
5277 -mpower9-minmax by default for -mcpu=power9.
5278 (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
5279 128-bit floating point.
5281 2017-01-20 Alan Modra <amodra@gmail.com>
5283 * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
5284 optimizing for size.
5286 2017-01-20 Alan Modra <amodra@gmail.com>
5289 * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
5290 for strcmp and strncmp from corresponding builtin decl.
5292 2017-01-19 Uros Bizjak <ubizjak@gmail.com>
5294 * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
5295 instead of i386/rtems-64.h.
5296 * config/i386/rtems-64.h: Remove.
5298 2017-01-19 Uros Bizjak <ubizjak@gmail.com>
5302 2013-11-05 Uros Bizjak <ubizjak@gmail.com>
5304 * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
5306 2017-01-19 Tamar Christina <tamar.christina@arm.com>
5308 * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
5309 Change int to HOST_WIDE_INT.
5310 * config/aarch64/aarch64-protos.h
5311 (aarch64_simd_gen_const_vector_dup): Likewise.
5312 * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
5314 2017-01-19 David Malcolm <dmalcolm@redhat.com>
5316 * langhooks-def.h (lhd_type_for_size): New decl.
5317 (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
5318 * langhooks.c (lhd_type_for_size): New function, taken from
5321 2017-01-19 Pat Haugen <pthaugen@us.ibm.com>
5323 * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
5324 define_bypass for CR latency.
5325 (power9-cracked-alu): Update bypass latency and remove power9-branch.
5326 (power9-alu2): Add define_bypass for CR latency.
5328 (power9-mul): Update insn latency.
5329 (power9-mul-compare): Update insn latency, bypass latency and remove
5332 2016-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5334 * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
5336 * config/aarch64/aarch64.md
5337 (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
5338 aarch64_nopcrelative_literal_loads.
5339 (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
5341 2017-01-19 Chenghua Xu <paul.hua.gm@gmail.com>
5343 * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
5345 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
5347 2017-01-19 Doug Gilmore <doug.gilmore@imgtec.com>
5350 * config.gcc (supported_defaults): Add lxc1-sxc1.
5351 (with_lxc1_sxc1): Add validation.
5352 (all_defaults): Add lxc1-sxc1.
5353 * config/mips/mips.opt (mlxc1-sxc1): New option.
5354 * gcc/config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
5356 (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
5357 __mips_no_lxc1_sxc1.
5358 (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
5359 * gcc/doc/invoke.texi (-mlxc1-sxc1): Document the new option.
5360 * doc/install.texi (--with-lxc1-sxc1): Document the new option.
5362 2017-01-19 Richard Biener <rguenther@suse.de>
5364 PR tree-optimization/72488
5365 * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
5366 sure to restore SSA info.
5367 * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
5369 2017-01-19 Richard Earnshaw <rearnsha@arm.com>
5371 PR rtl-optimization/79121
5372 * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
5373 of the inner type when shifting an extended value.
5375 2017-01-17 Jan Hubicka <hubicka@ucw.cz>
5378 * symtab.c (symtab_node::equal_address_to): Fix comparing of
5379 interposable aliases.
5381 2017-01-18 Peter Bergner <bergner@vnet.ibm.com>
5384 * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
5385 Use the evmergelohi instruction.
5386 (mov_si<mode>_e500_subreg4_2_le): Likewise.
5387 (mov_sitf_e500_subreg8_2_be): Likewise.
5388 (mov_sitf_e500_subreg12_2_le): Likewise.
5389 (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
5390 (mov_si<mode>_e500_subreg4_2_be): Likewise.
5391 (mov_sitf_e500_subreg8_2_le): Likewise.
5392 (mov_sitf_e500_subreg12_2_be): Likewise.
5394 2017-01-18 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5396 * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
5397 attribute from vecsimple to vecperm.
5398 (altivec_vbpermq2): Likewise.
5400 2017-01-18 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5403 * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
5405 2017-01-18 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5406 * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
5407 * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
5408 strcmp. Fix bug where comparison didn't stop with zero byte. Fix
5409 case where N arg is SIZE_MAX.
5410 * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
5411 (cmpstrsi): Add pattern.
5413 2017-01-18 Michael Meissner <meissner@linux.vnet.ibm.com>
5415 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
5416 __builtin_vec_revb builtins.
5417 * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
5418 built-in functions to support generation of the ISA 3.0 XXBR<x>
5419 vector byte reverse instructions.
5420 (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
5421 (P9V_BUILTIN_XXBRD_V2DI): Likewise.
5422 (P9V_BUILTIN_XXBRD_V2DF): Likewise.
5423 (P9V_BUILTIN_XXBGW_V4SI): Likewise.
5424 (P9V_BUILTIN_XXBGW_V4SF): Likewise.
5425 (P9V_BUILTIN_XXBGH_V8HI): Likewise.
5426 (P9V_BUILTIN_VEC_REVB): Likewise.
5427 * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
5428 generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
5429 (p9_xxbrq_v16qi): Likewise.
5430 (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
5431 (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
5432 (p9_xxbrh_v8hi): Likewise.
5433 * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
5434 * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
5435 vec_revb built-in functions.
5437 2017-01-18 Uros Bizjak <ubizjak@gmail.com>
5439 PR rtl-optimization/78952
5440 * config/i386/i386.md (any_extract): New code iterator.
5441 (*insvqi_2): Use any_extract for source operand.
5442 (*insvqi_3): Use any_shiftrt for source operand.
5444 2017-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5446 * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
5448 (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
5450 2017-01-18 Matthias Klose <doko@ubuntu.com>
5452 * doc/install.texi: Allow default for --with-target-bdw-gc-include.
5454 2016-01-18 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5456 * config/rs6000/altivec.h (vec_bperm): Change #define.
5457 * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
5458 (altivec_vbpermq2): New define_insn.
5459 (altivec_vbpermd): Likewise.
5460 * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
5462 (VBPERMD): Likewise.
5463 (VBPERM): New polymorphic function interface.
5464 * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
5465 Add entries for P9V_BUILTIN_VEC_VBPERM.
5466 * doc/extend.texi: Add interfaces for vec_bperm.
5468 2017-01-18 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5470 * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
5471 first letter of error messages.
5472 (s390_resolve_overloaded_builtin): Likewise.
5473 * config/s390/s390.c (s390_expand_builtin): Likewise.
5474 (s390_invalid_arg_for_unprototyped_fn): Likewise.
5475 (s390_valid_target_attribute_inner_p): Likewise.
5476 * config/s390/s390.md ("tabort"): Likewise.
5478 2017-01-18 Toma Tabacu <toma.tabacu@imgtec.com>
5480 * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
5481 (ISA_AVOID_DIV_HILO): New macro.
5482 (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
5483 (ISA_HAS_DDIV): Likewise.
5485 2017-01-18 Markus Trippelsdorf <markus@trippelsdorf.de>
5487 * doc/invoke.texi (fabi-version): Correct number of occurrences.
5489 2017-01-18 Markus Trippelsdorf <markus@trippelsdorf.de>
5491 * doc/invoke.texi (fabi-version): Spelling fix.
5493 2017-01-18 Markus Trippelsdorf <markus@trippelsdorf.de>
5496 * doc/invoke.texi (fabi-version): Mention mangling fix for
5499 2017-01-18 Markus Trippelsdorf <markus@trippelsdorf.de>
5502 * doc/invoke.texi (fabi-version): Document discriminator mangling.
5504 2017-01-17 Segher Boessenkool <segher@kernel.crashing.org>
5507 * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
5508 * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
5510 * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
5512 (stack_protect_test): Ditto.
5513 * config/rs6000/rs6000.opt (mstack-protector-guard=,
5514 mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
5516 * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
5517 Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
5518 -mstack-protector-guard-offset=.
5519 (RS/6000 and PowerPC Options): Ditto.
5521 2017-01-17 Uros Bizjak <ubizjak@gmail.com>
5523 * config/i386/i386.h (MASK_CLASS_P): New define.
5524 * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
5525 there are no registers from different register sets also when
5526 mask registers are used. Update function comment.
5527 * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
5528 to (*k/*r) and (*k/*km) alternatives.
5530 2017-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5532 * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
5533 * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
5534 (EH_RETURN_HANDLER_RTX): New define.
5535 * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
5536 Force frame pointer in EH return functions.
5537 (aarch64_expand_epilogue): Add barrier for eh_return.
5538 (aarch64_final_eh_return_addr): Remove.
5539 (aarch64_eh_return_handler_rtx): New function.
5540 * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
5542 (aarch64_eh_return_handler_rtx): New prototype.
5544 2017-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5546 * config/rs6000/altivec.h (vec_rlmi): New #define.
5547 (vec_vrlnm): Likewise.
5548 (vec_rlnm): Likewise.
5549 * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
5550 (UNSPEC_VRLNM): Likewise.
5551 (VIlong): New mode iterator.
5552 (altivec_vrl<VI_char>mi): New define_insn.
5553 (altivec_vrl<VI_char>nm): Likewise.
5554 * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
5557 (RLNM): New polymorphic function entry.
5558 (VRLWMI): New monomorphic function entry.
5560 (RLMI): New polymorphic function entry.
5561 * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
5562 new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
5563 * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
5566 2017-01-17 Jakub Jelinek <jakub@redhat.com>
5569 * dwarf2out.c (field_byte_offset): Restore the
5570 PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
5571 and DECL_FIELD_BIT_OFFSET. Use fold_build2 instead of build2 + fold.
5572 (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
5575 2017-01-17 Eric Botcazou <ebotcazou@adacore.com>
5578 * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
5580 2017-01-17 Jakub Jelinek <jakub@redhat.com>
5583 * dwarf2out.c (add_data_member_location_attribute): For constant
5584 offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
5585 instead of DW_AT_data_member_location, DW_AT_bit_offset and
5586 DW_AT_byte_size attributes.
5588 2017-01-17 Eric Botcazou <ebotcazou@adacore.com>
5590 * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
5591 after forcing to constant memory when the code model is medium.
5593 2017-01-17 Julia Koval <julia.koval@intel.com>
5596 * config/i386/avx512fintrin.h
5597 (_mm512_i32gather_ps): Change __addr type to void const*.
5598 (_mm512_mask_i32gather_ps): Ditto.
5599 (_mm512_i32gather_pd): Ditto.
5600 (_mm512_mask_i32gather_pd): Ditto.
5601 (_mm512_i64gather_ps): Ditto.
5602 (_mm512_mask_i64gather_ps): Ditto.
5603 (_mm512_i64gather_pd): Ditto.
5604 (_mm512_mask_i64gather_pd): Ditto.
5605 (_mm512_i32gather_epi32): Ditto.
5606 (_mm512_mask_i32gather_epi32): Ditto.
5607 (_mm512_i32gather_epi64): Ditto.
5608 (_mm512_mask_i32gather_epi64): Ditto.
5609 (_mm512_i64gather_epi32): Ditto.
5610 (_mm512_mask_i64gather_epi32): Ditto.
5611 (_mm512_i64gather_epi64): Ditto.
5612 (_mm512_mask_i64gather_epi64): Ditto.
5613 (_mm512_i32scatter_ps): Change __addr type to void*.
5614 (_mm512_mask_i32scatter_ps): Ditto.
5615 (_mm512_i32scatter_pd): Ditto.
5616 (_mm512_mask_i32scatter_pd): Ditto.
5617 (_mm512_i64scatter_ps): Ditto.
5618 (_mm512_mask_i64scatter_ps): Ditto.
5619 (_mm512_i64scatter_pd): Ditto.
5620 (_mm512_mask_i64scatter_pd): Ditto.
5621 (_mm512_i32scatter_epi32): Ditto.
5622 (_mm512_mask_i32scatter_epi32): Ditto.
5623 (_mm512_i32scatter_epi64): Ditto.
5624 (_mm512_mask_i32scatter_epi64): Ditto.
5625 (_mm512_i64scatter_epi32): Ditto.
5626 (_mm512_mask_i64scatter_epi32): Ditto.
5627 (_mm512_i64scatter_epi64): Ditto.
5628 (_mm512_mask_i64scatter_epi64): Ditto.
5629 * config/i386/avx512pfintrin.h
5630 (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
5631 (_mm512_mask_prefetch_i32gather_ps): Ditto.
5632 (_mm512_mask_prefetch_i64gather_pd): Ditto.
5633 (_mm512_mask_prefetch_i64gather_ps): Ditto.
5634 (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
5635 (_mm512_prefetch_i32scatter_ps): Ditto.
5636 (_mm512_mask_prefetch_i32scatter_pd): Ditto.
5637 (_mm512_mask_prefetch_i32scatter_ps): Ditto.
5638 (_mm512_prefetch_i64scatter_pd): Ditto.
5639 (_mm512_prefetch_i64scatter_ps): Ditto.
5640 (_mm512_mask_prefetch_i64scatter_pd): Ditto.
5641 (_mm512_mask_prefetch_i64scatter_ps): Ditto.
5642 * config/i386/avx512vlintrin.h
5643 (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
5644 (_mm_mmask_i32gather_ps): Ditto.
5645 (_mm256_mmask_i32gather_pd): Ditto.
5646 (_mm_mmask_i32gather_pd): Ditto.
5647 (_mm256_mmask_i64gather_ps): Ditto.
5648 (_mm_mmask_i64gather_ps): Ditto.
5649 (_mm256_mmask_i64gather_pd): Ditto.
5650 (_mm_mmask_i64gather_pd): Ditto.
5651 (_mm256_mmask_i32gather_epi32): Ditto.
5652 (_mm_mmask_i32gather_epi32): Ditto.
5653 (_mm256_mmask_i32gather_epi64): Ditto.
5654 (_mm_mmask_i32gather_epi64): Ditto.
5655 (_mm256_mmask_i64gather_epi32): Ditto.
5656 (_mm_mmask_i64gather_epi32): Ditto.
5657 (_mm256_mmask_i64gather_epi64): Ditto.
5658 (_mm_mmask_i64gather_epi64): Ditto.
5659 (_mm256_i32scatter_ps): Change __addr type to void*.
5660 (_mm256_mask_i32scatter_ps): Ditto.
5661 (_mm_i32scatter_ps): Ditto.
5662 (_mm_mask_i32scatter_ps): Ditto.
5663 (_mm256_i32scatter_pd): Ditto.
5664 (_mm256_mask_i32scatter_pd): Ditto.
5665 (_mm_i32scatter_pd): Ditto.
5666 (_mm_mask_i32scatter_pd): Ditto.
5667 (_mm256_i64scatter_ps): Ditto.
5668 (_mm256_mask_i64scatter_ps): Ditto.
5669 (_mm_i64scatter_ps): Ditto.
5670 (_mm_mask_i64scatter_ps): Ditto.
5671 (_mm256_i64scatter_pd): Ditto.
5672 (_mm256_mask_i64scatter_pd): Ditto.
5673 (_mm_i64scatter_pd): Ditto.
5674 (_mm_mask_i64scatter_pd): Ditto.
5675 (_mm256_i32scatter_epi32): Ditto.
5676 (_mm256_mask_i32scatter_epi32): Ditto.
5677 (_mm_i32scatter_epi32): Ditto.
5678 (_mm_mask_i32scatter_epi32): Ditto.
5679 (_mm256_i32scatter_epi64): Ditto.
5680 (_mm256_mask_i32scatter_epi64): Ditto.
5681 (_mm_i32scatter_epi64): Ditto.
5682 (_mm_mask_i32scatter_epi64): Ditto.
5683 (_mm256_i64scatter_epi32): Ditto.
5684 (_mm256_mask_i64scatter_epi32): Ditto.
5685 (_mm_i64scatter_epi32): Ditto.
5686 (_mm_mask_i64scatter_epi32): Ditto.
5687 (_mm256_i64scatter_epi64): Ditto.
5688 (_mm256_mask_i64scatter_epi64): Ditto.
5689 (_mm_i64scatter_epi64): Ditto.
5690 (_mm_mask_i64scatter_epi64): Ditto.
5691 * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
5692 (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
5693 (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
5694 (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
5695 (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
5696 (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
5697 (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
5698 (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
5699 (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
5700 (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
5701 (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
5702 (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
5703 (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
5704 (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
5705 (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
5706 (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
5707 (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
5708 (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
5709 (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
5710 (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
5711 (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
5712 (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
5713 (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
5714 (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
5715 (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
5716 (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
5717 (VOID_QI_V8DI_PCINT_INT_INT): Remove.
5718 (V16SF_V16SF_PCVOID_V16SI_HI_INT, V8DF_V8DF_PCVOID_V8SI_QI_INT)
5719 (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
5720 (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
5721 (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
5722 (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
5723 (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
5724 (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
5725 (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
5726 (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
5727 (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
5728 (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
5729 (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
5730 (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
5731 (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
5732 (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
5733 (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
5734 (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
5735 (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
5736 (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
5737 (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
5738 (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
5739 (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
5740 (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
5741 (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
5742 (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
5743 (VOID_QI_V8DI_PCVOID_INT_INT): Add.
5744 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
5745 definitions accordingly.
5747 2017-01-17 Kito Cheng <kito.cheng@gmail.com>
5748 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5751 * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
5754 2017-01-17 Vladimir Makarov <vmakarov@redhat.com>
5757 * ira-conflicts.c (ira_build_conflicts): Update total conflict
5758 hard regs for inner regno.
5760 2017-01-17 Martin Liska <mliska@suse.cz>
5763 * ipa-polymorphic-call.c (contains_type_p): Fix wrong
5764 assumption and add comment.
5766 2017-01-17 Nathan Sidwell <nathan@acm.org>
5768 * ipa-visibility.c (localize_node): New function, broken out of ...
5769 (function_and_variable_visibility): ... here. Call it.
5771 2017-01-17 Jan Hubicka <hubicka@ucw.cz>
5774 * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
5775 correctly set frequency of oudgoing edge.
5776 (duplicate_thread_path): Fix profile updating.
5778 2017-01-17 Jakub Jelinek <jakub@redhat.com>
5781 * configure.ac: Add GCC_BASE_VER.
5782 * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
5783 version from BASE-VER file.
5784 (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
5785 (gcc.o): Depend on $(BASEVER).
5786 * common.opt (dumpfullversion): New option.
5787 * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
5788 * doc/invoke.texi: Document -dumpfullversion.
5789 * doc/install.texi: Document --with-gcc-major-version-only.
5790 * configure: Regenerated.
5792 2017-01-17 Richard Biener <rguenther@suse.de>
5794 PR tree-optimization/71433
5795 * tree-vrp.c (register_new_assert_for): Merge same asserts
5796 on all incoming edges.
5797 (process_assert_insertions_for): Handle insertions at the
5800 2017-01-17 Gerald Pfeifer <gerald@pfeifer.com>
5802 * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
5803 * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
5805 2017-01-17 Kaz Kojima <kkojima@gcc.gnu.org>
5808 * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
5811 2017-01-17 Alan Modra <amodra@gmail.com>
5814 * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
5815 * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
5816 symbolic stack limit when pic.
5818 2017-01-16 Martin Sebor <msebor@redhat.com>
5820 PR tree-optimization/78608
5821 * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
5823 2017-01-16 Jeff Law <law@redhat.com>
5826 2016-12-02 Tadek Kijkowski <tkijkowski@gmail.com>
5827 * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
5828 for several include directories that may be relative to sysroot.
5829 * config/i386/x-mingw32 (gplus_includedir): Define.
5830 (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
5831 (native_system_includedir): Likewise.
5832 * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
5833 override if TARGET_SYSTEM_ROOT is defined.
5834 (NATIVE_SYSTEM_HEADER_DIR): Likewise.
5836 PR tree-optimization/79090
5837 PR tree-optimization/33562
5838 PR tree-optimization/61912
5839 PR tree-optimization/77485
5840 * tree-ssa-dse.c (compute_trims): Accept STMT argument. Dump STMT
5841 and computed trims into the dump file.
5843 2017-01-17 Uros Bizjak <ubizjak@gmail.com>
5845 * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
5847 2017-01-16 Jakub Jelinek <jakub@redhat.com>
5850 * gimplify.c (gimplify_init_constructor): If want_value and
5851 object == lhs, unshare lhs to avoid invalid tree sharing. Formatting
5855 * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
5856 sequence. Formatting fixes.
5857 (doloop_optimize): Formatting fixes.
5860 * gcc.c (debug_level_greater_than_spec_func): New function.
5861 (static_spec_functions): Add debug-level-gt spec function.
5862 (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
5864 * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
5865 * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
5866 * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
5867 gpubnames, ggnu-pubnames, gno-record-gcc-switches,
5868 grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
5869 gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
5871 2017-01-16 Uros Bizjak <ubizjak@gmail.com>
5873 * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
5874 QImode fixups to general and mask registers only.
5876 2017-01-16 Carl Love <cel@us.ibm.com>
5878 * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
5879 for built-in functions
5880 vector signed char vec_nabs (vector signed char)
5881 vector signed short vec_nabs (vector signed short)
5882 vector signed int vec_nabs (vector signed int)
5883 vector signed long long vec_nabs (vector signed long long)
5884 vector float vec_nabs (vector float)
5885 vector double vec_nabs (vector double)
5886 * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
5888 * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
5889 * config/rs6000/altivec.h: New define for vec_nabs built-in function.
5890 * doc/extend.texi: Update the documentation file for the new built-in
5893 2017-01-16 Martin Sebor <msebor@redhat.com>
5895 * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
5898 2017-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5900 * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
5901 UNSPEC_VSX__XXSPLTD to require special splat handling.
5903 2017-01-16 David Malcolm <dmalcolm@redhat.com>
5906 * system.h: Poison strndup.
5908 2017-01-16 Alan Modra <amodra@gmail.com>
5911 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
5914 2017-01-16 Georg-Johann Lay <avr@gjlay.de>
5916 * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
5918 2017-01-15 Uros Bizjak <ubizjak@gmail.com>
5920 * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
5921 call recog here. Assert that INSN_CODE (insn) is non-negative.
5923 2017-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5926 * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
5928 * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
5929 in the currently scheduled RTL fragment.
5931 2017-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5933 PR rtl-optimization/78751
5934 * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
5937 2017-01-14 Jeff Law <law@redhat.com>
5939 PR tree-optimization/79090
5940 * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
5941 variable length stores.
5942 (compute_trims): Delete dead assignment to *trim_tail.
5943 (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
5946 2017-01-14 Bernd Schmidt <bschmidt@redhat.com>
5948 PR rtl-optimization/78626
5949 PR rtl-optimization/78727
5950 * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
5951 of a block, and split such blocks after everything else is finished.
5953 2017-01-14 Alan Modra <amodra@gmail.com>
5956 * combine.c (recog_for_combine_1): Set INSN_CODE before calling
5957 target legitimate_combined_insn.
5958 * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
5959 (rs6000_legitimate_combined_insn): New function.
5960 * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
5962 (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
5963 (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
5964 (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
5966 2017-01-14 Gerald Pfeifer <gerald@pfeifer.com>
5968 * doc/frontends.texi (G++ and GCC): Remove references to Java.
5970 2017-01-13 Jeff Law <law@redhat.com>
5972 PR tree-optimization/33562
5973 PR tree-optimization/61912
5974 PR tree-optimization/77485
5975 * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
5977 (delete_dead_assignment): Likewise.
5978 (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
5979 statement to delete_dead_call and delete_dead_assignment.
5981 2017-01-13 David Malcolm <dmalcolm@redhat.com>
5984 * substring-locations.c (format_warning_va): Strengthen case 1 so
5985 that both endpoints of the substring must be within the format
5986 range for just the substring to be printed.
5988 2017-01-13 Uros Bizjak <ubizjak@gmail.com>
5990 * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
5991 * config/i386/i386.c (ix86_target_string): Add missing options
5992 to isa_opts and reorder options by implied ISAs. Rename isa_opts2 to
5993 isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
5994 flags_other and ix86_target_other to flags2_other. Display unknown
5996 (ix86_valid_target_attribute_inner_p): Add missing options and
5997 reorder options by implied ISAs, as in ix86_target_string.
5999 2017-01-13 Richard Sandiford <richard.sandiford@arm.com>
6001 * hash-table.h (hash_table::too_empty_p): New function.
6002 (hash_table::expand): Use it.
6003 (hash_table::traverse): Likewise.
6004 (hash_table::empty_slot): Use sizeof (value_type) instead of
6005 sizeof (PTR) to convert bytes to elements. Shrink the table
6006 if the current size is excessive for the current number of
6009 2017-01-13 Richard Sandiford <richard.sandiford@arm.com>
6011 * ira-costs.c (record_reg_classes): Break from the inner loop
6012 early once alt_fail is known to be true. Update outer loop
6013 handling accordingly.
6015 2017-01-13 Jeff Law <law@redhat.com>
6017 * tree-ssa-dse.c (decrement_count): New function.
6018 (increment_start_addr, maybe_trim_memstar_call): Likewise.
6019 (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
6020 when we know the partially dead statement is a mem* function.
6022 PR tree-optimization/61912
6023 PR tree-optimization/77485
6024 * tree-ssa-dse.c: Include expr.h.
6025 (maybe_trim_constructor_store): New function.
6026 (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
6028 PR tree-optimization/33562
6029 PR tree-optimization/61912
6030 PR tree-optimization/77485
6031 * doc/invoke.texi: Document new dse-max-object-size param.
6032 * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
6033 * tree-ssa-dse.c: Include params.h.
6034 (dse_store_status): New enum.
6035 (initialize_ao_ref_for_dse): New, partially extracted from
6037 (valid_ao_ref_for_dse, normalize_ref): New.
6038 (setup_live_bytes_from_ref, compute_trims): Likewise.
6039 (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
6040 (maybe_trim_partially_dead_store): Likewise.
6041 (maybe_trim_complex_store): Likewise.
6042 (dse_classify_store): Renamed from dse_possibly_dead_store_p.
6043 Track what bytes live from the original store. Return tri-state
6044 for dead, partially dead or live.
6045 (dse_dom_walker): Add constructor, destructor and new private members.
6046 (delete_dead_call, delete_dead_assignment): New extracted from
6048 (dse_optimize_stmt): Make a member of dse_dom_walker.
6049 Use initialize_ao_ref_for_dse.
6051 PR tree-optimization/33562
6052 PR tree-optimization/61912
6053 PR tree-optimization/77485
6054 * sbitmap.h (bitmap_count_bits): Prototype.
6055 (bitmap_clear_range, bitmap_set_range): Likewise.
6056 * sbitmap.c (bitmap_clear_range): New function.
6057 (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
6059 2017-01-13 Martin Liska <mliska@suse.cz>
6062 * function.c (set_cfun): Add new argument force.
6063 * function.h (set_cfun): Likewise.
6064 * ipa-inline-transform.c (inline_call): Use the function when
6065 strict alising from is dropped for function we inline to.
6067 2017-01-13 Richard Biener <rguenther@suse.de>
6069 * tree-pretty-print.c (dump_generic_node): Fix inverted condition
6070 for dumping GIMPLE INTEGER_CSTs.
6072 2017-01-13 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6074 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
6075 to 201112L since C++17.
6077 2017-01-13 Maxim Ostapenko <m.ostapenko@samsung.com>
6080 * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
6081 if -fsanitize=kernel-address is present.
6083 2017-01-13 Richard Biener <rguenther@suse.de>
6085 * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
6086 as _Literal ( type ) number in case usual suffixes do not
6087 preserve all information.
6089 2017-01-13 Richard Biener <rguenther@suse.de>
6091 PR tree-optimization/77283
6092 * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
6093 and ssa-iterators.h.
6094 (is_feasible_trace): Implement a cost model based on joiner
6097 2017-01-12 Michael Meissner <meissner@linux.vnet.ibm.com>
6100 * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
6101 char or short to __float128/_Float128 directly.
6103 2017-01-12 Martin Sebor <msebor@redhat.com>
6105 to -Wformat-overflow.
6106 * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
6107 (min_bytes_remaining): Same.
6108 (get_string_length): Same.
6109 (format_string): Same.
6110 (format_directive): Same.
6112 (pass_sprintf_length::handle_gimple_call): Same.
6114 2017-01-12 Jakub Jelinek <jakub@redhat.com>
6116 * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
6117 info.nowrite calls with no lhs that can't throw. Return bool
6118 whether gsi_remove has been called or not.
6119 (pass_sprintf_length::handle_gimple_call): Return bool whether
6120 try_substitute_return_value called gsi_remove. Formatting fix.
6121 (pass_sprintf_length::execute): Don't use gsi_remove if
6122 handle_gimple_call returned true.
6125 * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
6126 be removed due to side-effects, don't remove following barrier nor
6127 turn the successor edge into fallthru edge.
6129 2017-01-12 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6132 * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
6133 element-reversing loads and stores as not swappable.
6135 2017-01-12 Nathan Sidwell <nathan@acm.org>
6136 Nicolai Stange <nicstange@gmail.com>
6138 * combine.c (try_combine): Don't ignore result of overlap checking
6139 loop. Combine overlap & asm check into single loop.
6141 2017-01-12 Richard Biener <rguenther@suse.de>
6143 * tree-pretty-print.c (dump_generic_node): Provide -gimple
6144 variant for MEM_REF. Sanitize INTEGER_CST for -gimple.
6146 2017-01-12 Richard Biener <rguenther@suse.de>
6148 * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
6149 and TS_TARGET_OPTION directly derive from TS_BASE.
6150 * tree-core.h (tree_optimization_option): Derive from tree_base.
6151 (tree_target_option): Likewise.
6153 2017-01-11 Uros Bizjak <ubizjak@gmail.com>
6155 * config/i386/i386.c (memory_address_length): Increase len
6156 only when rip_relative_addr_p returns false.
6158 2017-01-11 Julia Koval <julia.koval@intel.com>
6160 * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
6161 (OPTION_MASK_ISA_SGX_SET): New.
6162 (ix86_handle_option): Handle OPT_msgx.
6163 * config.gcc: Added sgxintrin.h.
6164 * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
6165 * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
6166 * config/i386/i386.c (ix86_target_string): Add -msgx.
6168 (ix86_option_override_internal): Handle new options.
6169 (ix86_valid_target_attribute_inner_p): Add sgx.
6170 * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
6171 * config/i386/i386.opt: Add msgx.
6172 * config/i386/sgxintrin.h: New file.
6173 * config/i386/x86intrin.h: Add sgxintrin.h.
6175 2017-01-11 Jakub Jelinek <jakub@redhat.com>
6178 * fold-const.c (maybe_nonzero_address): Return 1 for function
6180 (tree_single_nonzero_warnv_p): Don't handle function local objects
6184 * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
6187 2017-01-11 David Malcolm <dmalcolm@redhat.com>
6190 * opts.c: Include "spellcheck.h"
6191 (struct string_fragment): New struct.
6192 (struct edit_distance_traits<const string_fragment &>): New
6194 (get_closest_sanitizer_option): New function.
6195 (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
6197 2017-01-11 Jakub Jelinek <jakub@redhat.com>
6199 * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
6201 (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
6202 DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
6203 (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
6204 (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
6205 for initial die_offset if dwarf_split_debug_info.
6206 (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
6207 initial next_die_offset if dwo_id is non-NULL. Don't emit padding
6209 (output_skeleton_debug_sections): Formatting fix. Use
6210 DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
6211 DWARF_COMPILE_UNIT_HEADER_SIZE. Don't emit padding.
6213 2017-01-11 Wilco Dijkstra <wdijkstr@arm.com>
6215 * config/arm/cortex-a53.md: Add bypasses for
6217 (cortex_a53_r2f): Only use for transfers.
6218 (cortex_a53_f2r): Likewise.
6219 (cortex_a53_r2f_cvt): Add reservation for conversions.
6220 (cortex_a53_f2r_cvt): Likewise.
6222 2017-01-11 Tamar Christina <tamar.christina@arm.com>
6224 * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
6225 to all inlined functions, change static to extern.
6227 2017-01-11 Christophe Lyon <christophe.lyon@linaro.org>
6230 * config/arm/arm.c (legitimize_pic_address): Handle reference to
6232 (arm_assemble_integer): Likewise.
6234 2017-01-11 Richard Earnshaw <rearnsha@arm.com>
6236 * config.gcc: Use new awk script to check CPU, FPU and architecture
6237 parameters for --with-... options.
6238 * config/arm/parsecpu.awk: New file
6239 * config/arm/arm-cpus.in: New file.
6240 * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
6242 * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
6244 * config/arm/t-arm: Update dependency rules.
6245 * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
6246 of processing .def files.
6247 * config/arm/genopt.sh: Deleted.
6248 * config/arm/gentune.sh: Deleted.
6249 * config/arm/arm-cores.def: Deleted.
6250 * config/arm/arm-arches.def: Deleted.
6251 * config/arm/arm-fpus.def: Deleted.
6252 * config/arm/arm-tune.md: Regenerated.
6253 * config/arm/arm-tables.opt: Regenerated.
6254 * config/arm/arm-cpu.h: New generated file.
6255 * config/arm/arm-cpu-data.h: New generated file.
6256 * config/arm/arm-cpu-cdata.h: New generated file.
6258 2017-01-11 Maxim Ostapenko <m.ostapenko@samsung.com>
6261 * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
6263 (input_varpool_node): Unpack dynamically_initialized bit.
6265 2017-01-11 Eric Botcazou <ebotcazou@adacore.com>
6267 PR rtl-optimization/79032
6268 * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
6269 the alignment of the adjusted memory reference against that of MODE,
6270 instead of the alignment of the original memory reference.
6272 2017-01-11 Martin Jambor <mjambor@suse.cz>
6274 * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
6276 * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
6277 decorated functions.
6279 2017-01-11 Richard Biener <rguenther@suse.de>
6281 * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
6282 set range/nonnull info for PHI results. Do not set it on
6283 stmts marked for removal.
6285 2017-01-10 Eric Botcazou <ebotcazou@adacore.com>
6287 * expr.c (store_field): In the bitfield case, fetch the return value
6288 from the registers before applying a single big-endian adjustment.
6289 Always do a final load for a BLKmode value not larger than a word.
6291 2017-01-10 David Malcolm <dmalcolm@redhat.com>
6294 * input.c (selftest::test_accessing_ordinary_linemaps): Verify
6295 that we correctly handle column numbers greater than
6296 LINE_MAP_MAX_COLUMN_NUMBER.
6298 2017-01-10 Martin Sebor <msebor@redhat.com>
6301 * gimple-ssa-sprintf.c (get_destination_size): Call
6302 {init,fini}object_sizes.
6303 * tree-object-size.c (addr_object_size): Adjust.
6304 (pass_through_call): Adjust.
6305 (pass_object_sizes::execute): Adjust.
6306 * tree-object-size.h (fini_object_sizes): Declare.
6308 2017-01-10 Martin Sebor <msebor@redhat.com>
6310 PR tree-optimization/78775
6311 * builtins.c (get_size_range): Move...
6312 * calls.c: ...to here.
6313 (alloc_max_size): Accept zero argument.
6314 (operand_signed_p): Remove.
6315 (maybe_warn_alloc_args_overflow): Call get_size_range.
6316 * calls.h (get_size_range): Declare.
6318 2017-01-10 Joe Seymour <joe.s@somniumtech.com>
6320 * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
6321 from TI's devices.csv file as of September 2016.
6322 * config/msp430/msp430.c (msp430_mcu_data): Likewise.
6324 2017-01-10 Sandra Loosemore <sandra@codesourcery.com>
6326 * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
6327 * doc/invoke.texi: Likewise.
6328 * doc/md.texi: Likewise.
6329 * doc/objc.texi: Likewise.
6331 2017-01-10 Joshua Conner <joshconner@google.com>
6333 * config/arm/fuchsia-elf.h: New file.
6334 * config/fuchsia.h: New file.
6335 * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
6336 (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
6338 * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
6340 2016-01-10 Richard Biener <rguenther@suse.de>
6342 PR tree-optimization/79034
6343 * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
6344 Propagate out degenerate PHIs in the joiner.
6346 2017-01-10 Martin Liska <mliska@suse.cz>
6348 * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
6349 (sort_congruence_classes_by_decl_uid): Likewise.
6350 (sort_congruence_class_groups_by_decl_uid): Likewise.
6351 (sem_item_optimizer::merge_classes): Sort class, groups in these
6352 classes and members in the groups by DECL_UID of declarations.
6353 This would make merge operations stable.
6355 2017-01-10 Martin Liska <mliska@suse.cz>
6357 * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
6358 usage of m_classes_vec.
6359 (sem_item_optimizer::~sem_item_optimizer): Likewise.
6360 (sem_item_optimizer::get_group_by_hash): Likewise.
6361 (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
6362 (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
6363 (sem_item_optimizer::verify_classes): Likewise.
6364 (sem_item_optimizer::process_cong_reduction): Likewise.
6365 (sem_item_optimizer::dump_cong_classes): Likewise.
6366 (sem_item_optimizer::merge_classes): Likewise.
6367 * ipa-icf.h (congruence_class_hash): Rename from
6368 congruence_class_group_hash. Remove declaration of m_classes_vec.
6370 2017-01-10 Andrew Senkevich <andrew.senkevich@intel.com>
6372 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
6373 OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
6374 * config.gcc: Add avx512vpopcntdqintrin.h.
6375 * config/i386/avx512vpopcntdqintrin.h: New.
6376 * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
6377 * config/i386/i386-builtin-types.def: Add new types.
6378 * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
6379 __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
6380 __builtin_ia32_vpopcountq_v8di_mask): New.
6381 * config/i386/i386-c.c (ix86_target_macros_internal): Define
6382 __AVX512VPOPCNTDQ__.
6383 * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
6384 (PTA_AVX512VPOPCNTDQ): Define.
6385 * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
6386 TARGET_AVX512VPOPCNTDQ_P): Define.
6387 * config/i386/i386.opt: Add mavx512vpopcntdq.
6388 * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
6389 * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
6391 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
6394 * predict.def (PRED_CALL): Set to 67.
6396 2017-01-09 Eric Botcazou <ebotcazou@adacore.com>
6398 * expr.c (store_field): In the bitfield case, if the value comes from
6399 a function call and is of an aggregate type returned in registers, do
6400 not modify the field mode; extract the value in all cases if the mode
6401 is BLKmode and the size is not larger than a word.
6403 2017-01-09 Dominique d'Humieres <dominiq@lps.ens.fr>
6406 * config/i386/cpuid.h: Fix undefined behavior.
6408 2017-01-04 Jeff Law <law@redhat.com>
6410 PR tree-optimization/79007
6411 PR tree-optimization/67955
6412 * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
6413 conservative for pt.null when flag_non_call_exceptions is on.
6415 2017-01-09 Jakub Jelinek <jakub@redhat.com>
6417 PR translation/79019
6418 PR translation/79020
6419 * params.def (PARAM_INLINE_MIN_SPEEDUP,
6420 PARAM_IPA_CP_SINGLE_CALL_PENALTY,
6421 PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
6423 * config/avr/avr.opt (maccumulate-args): Likewise.
6424 * config/msp430/msp430.opt (mwarn-mcu): Likewise.
6425 * common.opt (freport-bug): Likewise.
6426 * cif-code.def (CIF_FINAL_ERROR): Likewise.
6427 * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
6428 * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
6429 translatable string.
6430 * config/i386/i386.c (function_value_32): Likewise.
6431 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
6432 * config/msp430/msp430.c (msp430_option_override, msp430_attr):
6434 * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
6435 * common/config/msp430/msp430-common.c (msp430_handle_option):
6437 * symtab.c (symtab_node::verify_base): Likewise.
6438 * opts.c (set_debug_level): Likewise.
6439 * tree.c (verify_type_variant): Likewise. Fix typo in comment.
6440 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
6441 missing whitespace to translatable strings.
6442 * config/avr/avr.md (bswapsi2): Fix typo in comment.
6443 * config/sh/superh.h: Likewise.
6444 * config/i386/xopintrin.h: Likewise.
6445 * config/i386/znver1.md: Likewise.
6446 * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
6447 * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
6448 * double-int.h (struct double_int): Likewise.
6449 * double-int.c (div_and_round_double): Likewise.
6450 * wide-int.cc: Likewise.
6451 * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
6452 * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
6453 * cfgcleanup.c (crossjumps_occured): Renamed to ...
6454 (crossjumps_occurred): ... this.
6455 (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
6458 PR tree-optimization/78899
6459 * tree-if-conv.c (version_loop_for_if_conversion): Instead of
6460 returning bool return struct loop *, NULL for failure and the new
6462 (versionable_outer_loop_p): Don't version outer loop if it has
6463 dont_vectorized bit set.
6464 (tree_if_conversion): When versioning outer loop, ensure
6465 tree_if_conversion is performed also on the inner loop of the
6466 non-vectorizable outer loop copy.
6467 * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix. Fold
6468 LOOP_VECTORIZED in inner loop of the scalar outer loop and
6469 prevent vectorization of it.
6470 (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
6471 the outer loop vectorization of the non-scalar version is attempted
6472 before vectorization of the inner loop in scalar version. If
6473 outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
6474 vectorization of its inner loop.
6475 * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
6476 has 2 inner loops, rename also on edges from bb whose single pred
6477 is outer_loop->header. Fix typo in function comment.
6479 2017-01-09 Martin Sebor <msebor@redhat.com>
6482 * asan.c (asan_emit_stack_protection): Increase local buffer size
6483 to avoid snprintf truncation warning.
6485 2017-01-09 Andrew Pinski <apinski@cavium.com>
6487 * config/aarch64/aarch64-cores.def: Add thunderx2t99. Change vulcan
6488 to reference thunderx2t99 for the tuning structure
6489 * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
6491 (thunderx2t99_extra_costs): This.
6492 * config/aarch64/aarch64-tune.md: Regenerate.
6493 * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
6494 (vulcan_addrcost_table): This.
6495 (vulcan_regmove_cost): Rename to ...
6496 (thunderx2t99_regmove_cost): This.
6497 (vulcan_vector_cost): Rename to ...
6498 (thunderx2t99_vector_cost): this.
6499 (vulcan_branch_cost): Rename to ...
6500 (thunderx2t99_branch_cost): This.
6501 (vulcan_tunings): Rename to ...
6502 (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
6503 * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
6505 2017-01-09 Martin Jambor <mjambor@suse.cz>
6509 * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
6510 * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
6511 (propagate_vr_accross_jump_function): Use the above function for all
6512 value range computations for pass-through jump functions and type
6513 converasion from explicit value range values.
6514 (ipcp_propagate_stage): Do not attempt to deduce types of formal
6515 parameters from TYPE_ARG_TYPES.
6516 * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
6517 (ipa_write_node_info): Stream type of the actual argument.
6518 (ipa_read_node_info): Likewise. Also remove trailing whitespace.
6520 2017-01-09 Martin Liska <mliska@suse.cz>
6523 * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
6524 (lookup_compiler): Do not show error message with have_E.
6526 2017-01-09 Jakub Jelinek <jakub@redhat.com>
6528 PR tree-optimization/78938
6529 * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
6530 where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
6531 BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
6532 {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR. Formatting
6535 2017-01-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6537 * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
6540 2017-01-09 Richard Biener <rguenther@suse.de>
6542 PR tree-optimization/78997
6543 * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
6544 name condition properly.
6546 2017-01-09 Richard Biener <rguenther@suse.de>
6549 * dwarf2out.c (is_cxx): New overload with context.
6550 (is_naming_typedef_decl): Use it.
6552 2017-01-08 Sandra Loosemore <sandra@codesourcery.com>
6554 * invoke.texi (Option Summary): Correct spacing in option lists
6555 and add line breaks to fix over-long lines.
6557 2017-01-08 Sandra Loosemore <sandra@codesourcery.com>
6561 * extend.texi (Common Variable Attributes): Add xref to GCC
6562 Internals manual to explain mode attribute keywords.
6564 2017-01-08 Sandra Loosemore <sandra@codesourcery.com>
6567 * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
6568 and Preprocessor Options.
6569 (Options for Linking): Document -pthread here....
6570 (RS/6000 and PowerPC Options): ...not here.
6571 (Solaris 2 Options): ...or here.
6572 * doc/cppopts.texi: Document -pthread.
6574 2017-01-08 Martin Sebor <msebor@redhat.com>
6577 * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
6578 * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
6579 New member functions.
6580 (format_directive): Used them.
6582 (pass_sprintf_length::handle_gimple_call): Same.
6583 * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
6584 to avoid truncation for any argument.
6585 (extract_affine_mul): Same.
6586 * tree.c (get_file_function_name): Same.
6588 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
6591 * predict.def (PRED_INDIR_CALL): Set to 86.
6593 2017-01-07 Sandra Loosemore <sandra@codesourcery.com>
6595 PR preprocessor/54124
6596 * doc/cppopts.texi: Reformat -d subtable to list the full name
6597 of the options. Add cross-reference to the docs for the general
6598 compiler -d options.
6599 * doc/invoke.texi (Developer Options): Add cross-reference to the
6600 preprocessor-specific -d option documentation.
6602 2017-01-07 Sandra Loosemore <sandra@codesourcery.com>
6604 PR preprocessor/13498
6605 * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
6606 redudant material, and reflect new command-line options.
6607 (System Headers): Likewise.
6609 2017-01-07 Sandra Loosemore <sandra@codesourcery.com>
6611 * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
6612 -isystem, and -idirafter. Copy-edit.
6613 * doc/cppopts.texi: Copy-edit. Remove contradiction about
6614 default for -ftrack-macro-expansion. Delete obsolete and
6615 badly-formatted implementation details about -fdebug-cpp output.
6616 * doc/cppwarnopts.texi: Copy-edit.
6618 2017-01-07 David Malcolm <dmalcolm@redhat.com>
6621 * input.c (selftest::test_accessing_ordinary_linemaps): Verify
6622 that the transition from a max line width >= 1<<10 to narrower
6623 lines works correctly.
6625 2017-01-07 Alexandre Oliva <aoliva@redhat.com>
6627 * doc/options.texi (PerFunction): New.
6628 * opt-functions.awk (switch_flags): Map both Optimization and
6629 PerFunction to CL_OPTIMIZATION.
6630 * opth-gen.awk: Test for PerFunction flag along with
6632 * optc-save-gen.awk: Likewise. Introduce var_opt_hash and set
6633 it only when the latter is present. Skip those that don't in
6634 the hash function generator.
6635 * common.opt (fvar-tracking): Mark as PerFunction instead of
6637 (fvar-tracking-assignments): Likewise.
6638 (fvar-tracking-assignments-toggle): Likewise.
6639 (fvar-tracking-uninit): Likewise.
6641 2017-01-07 Jakub Jelinek <jakub@redhat.com>
6643 PR translation/79018
6644 * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
6647 2017-01-06 Mikael Pettersson <mikpelinux@gmail.com>
6650 * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
6651 * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
6652 TARGET_LONG_JUMP_TABLE_OFFSETS.
6653 * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
6654 * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
6655 * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
6656 (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
6657 * config/m68k/m68k.md (tablejump expander): Likewise.
6658 (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
6659 TARGET_LONG_JUMP_TABLE_OFFSETS.
6660 (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
6661 * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
6663 2017-01-06 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
6664 David Holsgrove <david.holsgrove@xilinx.com>
6666 * common/config/microblaze/microblaze-common.c
6667 (TARGET_EXCEPT_UNWIND_INFO): Remove.
6668 * config/microblaze/microblaze-protos.h (microblaze_eh_return):
6670 * config/microblaze/microblaze.c (microblaze_must_save_register)
6671 (microblaze_expand_epilogue, microblaze_return_addr): Handle
6673 (microblaze_eh_return): New function.
6674 * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
6675 (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
6676 (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
6677 * config/microblaze/microblaze.md (eh_return): New pattern.
6679 2017-01-06 Jakub Jelinek <jakub@redhat.com>
6681 * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
6682 GCC_DIAGNOSTIC_STRINGIFY): Define.
6684 * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
6686 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
6688 * config/arm/arm.md (<mcrr>): New.
6690 * config/arm/arm.c (arm_arch5te): New.
6691 (arm_option_override): Set arm_arch5te.
6692 (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
6694 * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
6695 (arm_mcrr_qualifiers): ... this. New.
6696 (MRRC_QUALIFIERS): Define to...
6697 (arm_mrrc_qualifiers): ... this. New.
6698 * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
6700 * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
6701 * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
6702 (MRRCI, mrrc, MRRC): New.
6703 * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
6704 VUNSPEC_MRRC2): New.
6706 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
6708 * config/arm/arm.md (<mcr>): New.
6710 * config/arm/arm.c (arm_coproc_builtin_available): Add
6711 support for mcr, mrc, mcr2 and mrc2.
6712 * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
6713 (arm_mcr_qualifiers): ... this. New.
6714 (MRC_QUALIFIERS): Define to ...
6715 (arm_mrc_qualifiers): ... this. New.
6716 (MCR_QUALIFIERS): Define to ...
6717 (arm_mcr_qualifiers): ... this. New.
6718 * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
6720 * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
6721 * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
6722 * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
6725 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
6727 * config/arm/arm.md (*ldc): New.
6731 * config/arm/arm.c (arm_coproc_builtin_available): Add
6732 support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
6733 (arm_coproc_ldc_stc_legitimate_address): New.
6734 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
6735 'qualifier_const_pointer'.
6736 (LDC_QUALIFIERS): Define to...
6737 (arm_ldc_qualifiers): ... this. New.
6738 (STC_QUALIFIERS): Define to...
6739 (arm_stc_qualifiers): ... this. New.
6740 * config/arm/arm-protos.h
6741 (arm_coproc_ldc_stc_legitimate_address): New.
6742 * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
6743 __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
6744 * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
6745 stc2, stcl, stc2l): New.
6746 * config/arm/constraints.md (Uz): New.
6747 * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
6748 * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
6749 VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
6750 VUNSPEC_STC2L): New.
6752 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
6754 * config/arm/arm.md (<cdp>): New.
6755 * config/arm/arm.c (neon_const_bounds): Rename this ...
6756 (arm_const_bounds): ... this.
6757 (arm_coproc_builtin_available): New.
6758 * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
6759 (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
6760 (CDP_QUALIFIERS): Define to...
6761 (arm_cdp_qualifiers): ... this. New.
6763 (arm_expand_builtin_args): Add case for 6 arguments.
6764 * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
6765 (arm_const_bounds): ... this.
6766 (arm_coproc_builtin_available): New.
6767 * config/arm/arm_acle.h (__arm_cdp): New.
6769 * config/arm/arm_acle_builtins.def (cdp): New.
6771 * config/arm/iterators.md (CDPI,CDP,cdp): New.
6772 * config/arm/neon.md: Rename all 'neon_const_bounds' to
6774 * config/arm/types.md (coproc): New.
6775 * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
6776 * gcc/doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
6777 * gcc/doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
6778 arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
6780 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
6782 * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
6783 (UBINOP_QUALIFIERS): New.
6785 (acle_builtin_data): New. Change comment.
6786 (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
6787 ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
6788 ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
6789 arm_acle_builtins.def.
6790 (ARM_BUILTIN_ACLE_PATTERN_START): Define.
6791 (arm_init_acle_builtins): New.
6792 (CRC32_BUILTIN): Remove.
6793 (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
6794 crc32cb, crc32ch and crc32cw.
6795 (arm_init_crc32_builtins): Remove.
6796 (arm_init_builtins): Use arm_init_acle_builtins rather
6797 than arm_init_crc32_builtins.
6798 (arm_expand_acle_builtin): New.
6799 (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
6800 * config/arm/arm_acle_builtins.def: New.
6802 2017-01-06 Andre Vieira <andre.simoesdiasvieira@arm.com>
6804 * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
6805 (arm_builtin_datum): ... this.
6806 (arm_init_neon_builtin): Rename to ...
6807 (arm_init_builtin): ... this. Add a new parameters PREFIX
6808 and USE_SIG_IN_NAME.
6809 (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
6810 'arm_init_builtin'. Replace type 'neon_builtin_datum' with
6811 'arm_builtin_datum'.
6812 (arm_init_vfp_builtins): Likewise.
6813 (builtin_arg): Rename enum's replacing 'NEON_ARG' with
6814 'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
6815 (arm_expand_neon_args): Rename to ...
6816 (arm_expand_builtin_args): ... this. Rename builtin_arg
6817 enum values and differentiate between ARG_BUILTIN_MEMORY
6818 and ARG_BUILTIN_NEON_MEMORY.
6819 (arm_expand_neon_builtin_1): Rename to ...
6820 (arm_expand_builtin_1): ... this. Rename builtin_arg enum
6821 values, arm_expand_builtin_args and add bool parameter NEON.
6822 (arm_expand_neon_builtin): Use arm_expand_builtin_1.
6823 (arm_expand_vfp_builtin): Likewise.
6824 (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
6826 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
6829 * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
6830 * predict.c (tree_estimate_probability_bb): Reverse direction of
6831 polymorphic call predictor.
6833 2017-01-06 David Malcolm <dmalcolm@redhat.com>
6835 * passes.c (execute_one_pass): Split out pass-skipping logic into...
6836 (determine_pass_name_match): ...this new function and...
6837 (should_skip_pass_p): ...this new function.
6839 2017-01-06 Nathan Sidwell <nathan@acm.org>
6841 * ipa-visibility.c (function_and_variable_visibility): Reformat
6842 comments and long lines. Remove extrneous if.
6843 * symtab.c (symtab_node::make_decl_local): Fix code format.
6844 (symtab_node::set_section_for_node): Fix comment typo.
6846 2017-01-06 Martin Liska <mliska@suse.cz>
6849 * lra-constraints.c: Rename invariant to lra_invariant.
6850 * predict.c (set_even_probabilities): Initialize e to NULL.
6852 2017-01-05 Martin Sebor <msebor@redhat.com>
6854 PR tree-optimization/78910
6855 * gimple-ssa-sprintf.c (tree_digits): Add an argument.
6856 (format_integer): Correct off-by-one error in the handling
6857 of precision with negative numbers in signed conversions..
6859 2017-01-05 Eric Botcazou <ebotcazou@adacore.com>
6861 * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
6863 2017-01-05 Jakub Jelinek <jakub@redhat.com>
6865 PR tree-optimization/71016
6866 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
6867 factor_out_conditional_conversion. Formatting fix.
6868 (factor_out_conditional_conversion): Add cond_stmt argument.
6869 If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
6870 cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
6873 2017-01-05 David Malcolm <dmalcolm@redhat.com>
6875 * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
6876 read-rtl-function.o, and selftest-rtl.o.
6877 * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
6878 (selftest::aarch64_test_loading_full_dump): New function.
6879 (selftest::aarch64_run_selftests): New function.
6880 (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
6881 selftest::aarch64_run_selftests.
6882 * config/i386/i386.c
6883 (selftest::ix86_test_loading_dump_fragment_1): New function.
6884 (selftest::ix86_test_loading_call_insn): New function.
6885 (selftest::ix86_test_loading_full_dump): New function.
6886 (selftest::ix86_test_loading_unspec): New function.
6887 (selftest::ix86_run_selftests): Call the new functions.
6888 * emit-rtl.c (maybe_set_max_label_num): New function.
6889 * emit-rtl.h (maybe_set_max_label_num): New decl.
6890 * function.c (instantiate_decls): Guard call to
6891 instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
6892 * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
6894 * gensupport.c (gen_reader::gen_reader): Pass "false"
6895 for new "compact" param of rtx_reader.
6896 * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
6897 rather than an empty string for NULL strings.
6898 * read-md.c: Potentially include config.h rather than bconfig.h.
6899 Wrap include of errors.h with #ifdef GENERATOR_FILE.
6900 (have_error): New global, copied from errors.c.
6901 (md_reader::read_name): Rename to...
6902 (md_reader::read_name_1): ...this, adding "out_loc" param,
6903 and converting "missing name or number" to returning false, rather
6905 (md_reader::read_name): Reimplement in terms of read_name_1.
6906 (md_reader::read_name_or_nil): New function.
6907 (md_reader::read_string): Handle "(nil)" by returning NULL.
6908 (md_reader::md_reader): Add new param "compact".
6909 (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
6910 (md_reader::read_file): New method.
6911 * read-md.h (md_reader::md_reader): Add new param "compact".
6912 (md_reader::read_file): New method.
6913 (md_reader::is_compact): New accessor.
6914 (md_reader::read_name): Convert return type from void to file_location.
6915 (md_reader::read_name_or_nil): New decl.
6916 (md_reader::read_name_1): New decl.
6917 (md_reader::m_compact): New field.
6918 (noop_reader::noop_reader): Pass "false" for new "compact" param
6920 (rtx_reader::rtx_reader): Add new "compact" param.
6921 (rtx_reader::read_rtx_operand): Make virtual and convert return
6922 type from void to rtx.
6923 (rtx_reader::read_until): New decl.
6924 (rtx_reader::handle_any_trailing_information): New virtual function.
6925 (rtx_reader::postprocess): New virtual function.
6926 (rtx_reader::finalize_string): New virtual function.
6927 (rtx_reader::m_in_call_function_usage): New field.
6928 (rtx_reader::m_reuse_rtx_by_id): New field.
6929 * read-rtl-function.c: New file.
6930 * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
6931 * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
6932 (selftest::verify_three_block_rtl_cfg): New decl.
6933 * read-rtl-function.h: New file.
6934 * read-rtl.c: Potentially include config.h rather than bconfig.h.
6935 For host, include function.h, memmodel.h, and emit-rtl.h.
6936 (one_time_initialization): New function.
6937 (struct compact_insn_name): New struct.
6938 (compact_insn_names): New array.
6939 (find_code): Handle insn codes in compact dumps.
6940 (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
6941 (bind_subst_iter_and_attr): Likewise.
6942 (add_condition_to_string): Likewise.
6943 (add_condition_to_rtx): Likewise.
6944 (apply_attribute_uses): Likewise.
6945 (add_current_iterators): Likewise.
6946 (apply_iterators): Likewise.
6947 (initialize_iterators): Guard usage of apply_subst_iterator with
6948 #ifdef GENERATOR_FILE.
6949 (read_conditions): Wrap with #ifdef GENERATOR_FILE.
6950 (md_reader::read_mapping): Likewise.
6951 (add_define_attr_for_define_subst): Likewise.
6952 (add_define_subst_attr): Likewise.
6953 (read_subst_mapping): Likewise.
6954 (check_code_iterator): Likewise.
6955 (rtx_reader::read_rtx): Likewise. Move one-time initialization
6957 (one_time_initialization): New function.
6958 (rtx_reader::read_until): New method.
6959 (read_flags): New function.
6960 (parse_reg_note_name): New function.
6961 (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
6962 Handle reuse_rtx ids.
6963 Wrap iterator lookup within #ifdef GENERATOR_FILE.
6964 Add parsing support for RTL dumps, mirroring the special-cases in
6965 print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
6966 values, and calling handle_any_trailing_information.
6967 (rtx_reader::read_rtx_operand): Convert return type from void
6968 to rtx, returning return_rtx. Handle case 'e'. Call
6969 finalize_string on XSTR and XTMPL fields.
6970 (rtx_reader::read_nested_rtx): Handle dumps in which trailing
6971 "(nil)" values were omitted. Call the postprocess vfunc on the
6973 (rtx_reader::rtx_reader): Add new "compact" param and pass to base
6974 class ctor. Initialize m_in_call_function_usage. Call
6975 one_time_initialization.
6976 * rtl-tests.c (selftest::test_uncond_jump): Call
6977 set_new_first_and_last_insn.
6978 * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
6979 * selftest-rtl.c: New file.
6980 * selftest-rtl.h (class selftest::rtl_dump_test): New class.
6981 (selftest::get_insn_by_uid): New decl.
6982 * selftest-run-tests.c (selftest::run_tests): Call
6983 read_rtl_function_c_tests.
6984 * selftest.h (selftest::read_rtl_function_c_tests): New decl.
6985 * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
6988 2017-01-05 Uros Bizjak <ubizjak@gmail.com>
6990 * config/i386/i386.md (*testqi_ext_3): No need to handle memory
6991 operands in a special way. Assert that pos+len <= mode precision.
6993 2017-01-05 Jakub Jelinek <jakub@redhat.com>
6995 * common.opt (fvect-cost-model): Remove RejectNegative flag, use
6996 3 argument Alias with unlimited for the negative form.
6997 (fno-vect-cost-model): Removed.
6999 2017-01-05 Martin Liska <mliska@suse.cz>
7001 * hsa-gen.c (gen_hsa_divmod): New function.
7002 (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
7004 2017-01-05 Martin Liska <mliska@suse.cz>
7007 * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
7010 2017-01-05 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7012 * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
7013 small constant length operands.
7015 2017-01-05 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7017 * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
7018 between loop iterations.
7020 2017-01-05 Martin Liska <mliska@suse.cz>
7023 * gimplify.c (gimplify_decl_expr): Compare to
7024 asan_poisoned_variables instread of checking flags.
7025 (gimplify_target_expr): Likewise.
7026 (gimplify_expr): Likewise.
7027 (gimplify_function_tree): Conditionally initialize
7028 asan_poisoned_variables.
7030 2017-01-04 Jeff Law <law@redhat.com>
7032 PR tree-optimizatin/78812
7033 * rtl.h (contains_mem_rtx_p): Prototype.
7034 * ifcvt.c (containts_mem_rtx_p): Move from here to...
7035 * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
7036 * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
7037 and prune MEMs that are not at the toplevel of a SET_SRC rtx. Look
7038 through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
7040 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
7042 * input.c (assert_char_at_range): Default-initialize actual_range.
7044 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
7046 * df-scan.c (df_ref_create_structure): Make regno unsigned,
7047 to match the caller.
7049 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
7051 * cfgexpand.c (expand_gimple_basic_block): Disregard debug
7052 insns after final jump in test to emit dummy move.
7054 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
7056 * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
7057 * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
7059 2017-01-04 Alexandre Oliva <aoliva@redhat.com>
7061 * multiple_target.c (create_dispatcher_calls): Init e_next.
7062 * tree-ssa-loop-split.c (split_loop): Init border.
7063 * tree-vect-loop.c (vect_determine_vectorization_factor): Init
7066 2017-01-04 Michael Meissner <meissner@linux.vnet.ibm.com>
7071 * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
7072 (altivec_register_operand): Do not return true if the operand
7073 contains a SUBREG mixing SImode and SFmode.
7074 (vsx_register_operand): Likewise.
7075 (vsx_reg_sfsubreg_ok): New predicate.
7076 (vfloat_operand): Do not return true if the operand contains a
7077 SUBREG mixing SImode and SFmode.
7078 (vint_operand): Likewise.
7079 (vlogical_operand): Likewise.
7080 (gpc_reg_operand): Likewise.
7081 (int_reg_operand): Likewise.
7082 * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
7083 * config/rs6000/rs6000.c (valid_sf_si_move): New function to
7084 determine if a MOVSI or MOVSF operation contains SUBREGs that mix
7086 (rs6000_emit_move_si_sf_subreg): New helper function.
7087 (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
7088 fixup SUBREGs involving SImode and SFmode.
7089 * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
7090 numbers for the new peephole2 optimization.
7091 (peephole2 for SFmode unions): New peephole2 to optimize cases in
7092 the GLIBC math library that do AND/IOR/XOR operations on single
7093 precision floating point.
7094 * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
7095 target macros to say whether we need to avoid SUBREGs mixing
7097 (TARGET_ALLOW_SF_SUBREG): Likewise.
7098 * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
7099 (UNSPEC_SI_FROM_SF): Likewise.
7100 (iorxor): Change spacing.
7101 (and_ior_xor): New iterator for AND, IOR, and XOR.
7102 (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
7103 (movdi_from_sf_zero_ext): Likewise.
7104 (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
7105 instead of gpc_reg_operand. Add SImode/SFmode SUBREG support.
7106 (movsf_from_si): New insn for SImode/SFmode SUBREG support.
7107 (fma<mode>4): Use gpc_reg_operand instead of register_operand.
7108 (fms<mode>4): Likewise.
7109 (fnma<mode>4): Likewise.
7110 (fnms<mode>4): Likewise.
7111 (nfma<mode>4): Likewise.
7112 (nfms<mode>4): Likewise.
7114 2017-01-04 Marek Polacek <polacek@redhat.com>
7117 * doc/invoke.texi: Document -Wpointer-compare.
7119 2017-01-04 Jakub Jelinek <jakub@redhat.com>
7121 * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
7124 * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
7125 descriptions for -gdwarf-5 and emit them as uleb128 instead of
7128 2017-01-04 Kelvin Nilsen <kelvin@gcc.gnu.org>
7131 * doc/sourcebuild.texi (PowerPC-specific attributes): Add
7132 documentation of the powerpc_popcntb_ok attribute.
7133 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
7134 code to issue warning messages if a requested CPU configuration is
7135 not supported by the binary (assembler and loader) toolchain.
7136 (spe_init_builtins): Add two assertions to prevent ICE if attempt is
7137 made to define a built-in function that has been disabled.
7138 (paired_init_builtins): Add assertion to prevent ICE if attempt is
7139 made to define a built-in function that has been disabled.
7140 (altivec_init_builtins): Add comment explaining why definition
7141 of the DST built-in functions is not preceded by an assertion
7142 check. Add assertions to prevent ICE if attempts are made to
7143 define an altivec predicate or an abs* built-in function that has
7145 (htm_init_builtins): Add comment explaining why definition of the
7146 htm built-in functions is not preceded by an assertion check.
7148 2017-01-04 Jeff Law <law@redhat.com>
7150 PR tree-optimizatin/67955
7151 * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
7152 Allow any SSA_VAR_P as the base objects. Use integer_zerop. Verify
7153 the points-to solution does not include pt_null. Use DECL_PT_UID
7156 2017-01-04 Uros Bizjak <ubizjak@gmail.com>
7158 * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
7159 Use gen_int_mode instead of gen_lopwart for const_int operands.
7161 2017-01-04 Jakub Jelinek <jakub@redhat.com>
7163 PR tree-optimization/71563
7164 * match.pd: Simplify X << Y into X if Y is known to be 0 or
7165 out of range value - has low bits known to be zero.
7167 2017-01-04 Alan Modra <amodra@gmail.com>
7169 * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
7170 * configure: Regenerate.
7171 * config.in: Regenerate.
7173 2017-01-04 Jakub Jelinek <jakub@redhat.com>
7176 * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
7177 a substring of the message, but strcmp with the whole message. Ifdef
7178 ENABLE_NLS, translate the message first using dgettext.
7180 2017-01-03 Jeff Law <law@redhat.com>
7182 PR tree-optimizatin/78856
7183 * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
7184 (mark_threaded_blocks): Remove code to truncate thread paths that
7185 cross multiple loop headers. Instead invalidate the cached loop
7186 iteration information and handle case of a thread path walking
7187 into an irreducible region.
7189 2017-01-03 Michael Meissner <meissner@linux.vnet.ibm.com>
7192 * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
7193 assertions. Add support for doing the signbit if the IEEE 128-bit
7194 floating point value is in a GPR.
7195 * config/rs6000/rs6000.md (Fsignbit): Delete.
7196 (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
7197 Update the length attribute if the value is in a GPR.
7198 (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
7199 the sign or zero extension instruction, since the value is always 0/1.
7200 (signbit<mode>2_dm2): Delete using <Fsignbit>.
7203 * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
7204 extracting SImode to a GPR register so that we can generate a
7205 store, limit the vector to be in a traditional Altivec register
7206 for the vextuwrx instruction.
7208 2017-01-03 Ian Lance Taylor <iant@google.com>
7210 * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
7212 2017-01-03 Martin Sebor <msebor@redhat.com>
7214 PR tree-optimization/78696
7215 * gimple-ssa-sprintf.c (format_floating): Correct handling of
7216 precision. Use MPFR for %f for greater fidelity. Correct handling
7218 (pass_sprintf_length::compute_format_length): Set width and precision
7219 specified by asrerisk to void_node for vararg functions.
7220 (try_substitute_return_value): Adjust dump output.
7222 2017-01-03 David Edelsohn <dje.gcc@gmail.com>
7224 * doc/invoke.texi (RS6000 options): LRA is enabled by default.
7226 2017-01-03 Eric Botcazou <ebotcazou@adacore.com>
7228 * doc/invoke.texi (SPARC options): Document -mlra as the default.
7229 * config/sparc/sparc.c (sparc_option_override): Force LRA unless
7230 -mlra/-mno-lra was passed to the compiler.
7232 2017-01-03 James Cowgill <James.Cowgill@imgtec.com>
7234 PR rtl-optimization/65618
7235 * emit-rtl.c (try_split): Move initialization of "before" and
7236 "after" to just before the call to emit_insn_after_setloc.
7238 2017-01-03 Gerald Pfeifer <gerald@pfeifer.com>
7240 * doc/md.texi (Standard Names): Remove reference to Java frontend.
7242 2017-01-03 Pierre-Marie de Rodat <derodat@adacore.com>
7244 * dwarf2out.c (gen_enumeration_type_die): When
7245 -gno-strict-dwarf, add a DW_AT_encoding attribute.
7247 2017-01-03 Jakub Jelinek <jakub@redhat.com>
7249 PR tree-optimization/78965
7250 * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
7251 Change first argument from const call_info & to call_info &. For %n
7252 set info.nowrite to false.
7255 * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
7256 possibly throwing calls.
7258 * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
7259 and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
7260 and fns handling, rather than in a separate case SSA_NAME.
7262 2017-01-02 Jeff Law <law@redhat.com>
7264 * config/darwin-driver.c (darwin_driver_init): Const-correctness
7265 fixes for first_period and second_period variables.
7267 2017-01-02 Uros Bizjak <ubizjak@gmail.com>
7270 * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
7271 (*insvqi_1): New insn pattern.
7272 (*insvqi_1_mem_rex64): Ditto.
7274 (*insvqi_3): Rename from *insvqi.
7276 (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
7278 2017-01-02 Gerald Pfeifer <gerald@pfeifer.com>
7280 * doc/cfg.texi (Edges): Remove reference to Java.
7281 (Maintaining the CFG): Ditto.
7283 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
7286 * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
7287 transparent aliases.
7289 2017-01-01 Jan Hubicka <hubicka@ucw.cz>
7292 * predict.def (PRED_CALL): Update hitrate.
7293 (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
7294 * predict.c (tree_estimate_probability_bb): Split CALL predictor
7295 into direct/indirect/polymorphic variants.
7297 2017-01-01 Jakub Jelinek <jakub@redhat.com>
7299 Update copyright years.
7301 * gcc.c (process_command): Update copyright notice dates.
7302 * gcov-dump.c (print_version): Ditto.
7303 * gcov.c (print_version): Ditto.
7304 * gcov-tool.c (print_version): Ditto.
7305 * gengtype.c (create_file): Ditto.
7306 * doc/cpp.texi: Bump @copying's copyright year.
7307 * doc/cppinternals.texi: Ditto.
7308 * doc/gcc.texi: Ditto.
7309 * doc/gccint.texi: Ditto.
7310 * doc/gcov.texi: Ditto.
7311 * doc/install.texi: Ditto.
7312 * doc/invoke.texi: Ditto.
7314 Copyright (C) 2017 Free Software Foundation, Inc.
7316 Copying and distribution of this file, with or without modification,
7317 are permitted in any medium without royalty provided the copyright
7318 notice and this notice are preserved.