re PR sanitizer/79341 (Many Asan tests fail on s390)

PR sanitizer/79341
* configure.tgt (s390*-*-linux*): Don't disable libsanitizer on
s390-linux 31-bit.
* sanitizer_common/sanitizer_internal_defs.h: Cherry-pick upstream
r294793.
* sanitizer_common/sanitizer_common_interceptors.inc: Cherry-pick
upstream r294790.
* sanitizer_common/sanitizer_linux_s390.cc: Cherry-pick upstream
r294799.

From-SVN: r245350

* doc/xml/manual/profile_mode.xml: Update a paper reference.

From-SVN: r245349

makefile.texi (profiledbootstrap): Refer to the installation instructions only in textual form.

* doc/makefile.texi (profiledbootstrap): Refer to the
installation instructions only in textual form.

From-SVN: r245348

re PR target/79295 (gcc.target/powerpc/bcd-3.c fails starting with r244942)

2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>

PR target/79295
* config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.

Applying patch suggested by Meissner.

From-SVN: r245345

re PR c++/79457 (Segmentation fault in templated decltype evaluation)

PR c++/79457
* g++.dg/cpp0x/pr79457.C: New test.

From-SVN: r245344

PR c++/78908 - template ops and bitfields

* tree.c (build_min_non_dep): Use unlowered_expr_type.

From-SVN: r245343

PR c++/78897 - constexpr union

* constexpr.c (cxx_eval_store_expression): A store to a union member
erases a previous store to another member.

From-SVN: r245341

PR c++/71285 - member of fold-expression

* semantics.c (finish_unary_fold_expr)
(finish_binary_fold_expr): Use null type for fold-expressions.

From-SVN: r245340

PR c++/79401 - protected inherited constructor

* call.c (enforce_access): For inheriting constructor, find a base
binfo in the path we already have.

From-SVN: r245339

testsuite, rs6000: fold-vec-mult-longlong.c

On 32-bit we of course do not generate mulld insns.  This patch makes
the testcase only do the scan-assembler on 64-bit targets.

gcc/testsuite/
* gcc.target/powerpc/fold-vec-mult-longlong.c: Don't do the
scan-assembler unless lp64.

From-SVN: r245338

testsuite, rs6000: Don't xfail 32-bit (PR66612)

-m32 works fine, only 64-bit still fails.

gcc/testsuite/
PR tree-optimization/66612
* gcc.target/powerpc/20050830-1.c: Don't xfail on 32-bit.

From-SVN: r245337

testsuite, rs6000: Don't xfail Wtrampolines.c on -m32

This test does not fail with -m32, so we shouldn't xfail it there.

gcc/testsuite/
* gcc.dg/Wtrampolines.c: Do the xfail for only 64-bit powerpc instead
of all powerpc.

From-SVN: r245336

re PR c++/79184 (-Wint-in-bool-context triggered erroneously in template parameter)

PR c++/79184
* cvt.c (ocp_convert): Add a sentinel against -Wint-in-bool-context
if warnings shouldn't be given.

* g++.dg/warn/Wint-in-bool-context-1.C: New.

From-SVN: r245335

re PR c++/79435 (ICE on invalid C++ code (with member access into an incomplete type) on x86_64-linux-gnu: Segmentation fault)

PR c++/79435
* pt.c (type_dependent_expression_p): Check if the expression type
is null.

* g++.dg/cpp1y/pr79435.C: New.

From-SVN: r245334

install.texi (Specific): Use https for blackfin.uclinux.org.

* doc/install.texi (Specific): Use https for blackfin.uclinux.org.
(Specific): Update mingw-w64 reference.
(Binaries): Ditto.
(Specific): Remove broken link to Renesas RX processor.

From-SVN: r245332

toplev.c (process_options): Do not mention obsolete graphite options when...

2017-02-10  Richard Biener  <rguenther@suse.de>

* toplev.c (process_options): Do not mention obsolete graphite
options when printing sorry message about missing graphite support.
Mention -floop-nest-optimize.

From-SVN: r245331

[ARM,AArch64] more poly64 intrinsics and tests

2017-02-06  Christophe Lyon  <christophe.lyon@linaro.org>

gcc/
* config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
(vtst_p16): Likewise.
(vtstq_p8): Likewise.
(vtstq_p16): Likewise.
(vtst_p64): New.
(vtstq_p64): Likewise.
* config/arm/arm_neon.h (vgetq_lane_p64): New.
(vset_lane_p64): New.
(vsetq_lane_p64): New.

2017-02-06  Christophe Lyon  <christophe.lyon@linaro.org>

gcc/testsuite/
* gcc.target/aarch64/advsimd-intrinsics/p64_p128.c
(vget_lane_expected, vset_lane_expected, vtst_expected_poly64x1):
New.
(vmov_n_expected0, vmov_n_expected1, vmov_n_expected2)
(expected_vld_st2_0, expected_vld_st2_1, expected_vld_st3_0)
(expected_vld_st3_1, expected_vld_st3_2, expected_vld_st4_0)
(expected_vld_st4_1, expected_vld_st4_2, expected_vld_st4_3)
(vtst_expected_poly64x2): Move to aarch64-only section.
(vget_lane_p64, vgetq_lane_p64, vset_lane_p64, vsetq_lane_p64)
(vtst_p64, vtstq_p64): New tests.

From-SVN: r245328

re PR c++/71737 (ICE following 2x pack expansion in non-pack with template alias)

/cp
2017-02-10  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/71737
* pt.c (tsubst_decl): Don't try to preserve a typedef that names
an error_mark_node as type.

/testsuite
2017-02-10  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/71737
* g++.dg/cpp0x/pr71737.C: New.

From-SVN: r245327

gimple-parser.c (c_parser_gimple_binary_expression): Avoid building IL when arguments are error_mark_node.

2017-02-10  Prasad Ghangal  <prasad.ghangal@gmail.com>
Richard Biener  <rguenther@suse.de>

* gimple-parser.c (c_parser_gimple_binary_expression): Avoid
building IL when arguments are error_mark_node.
(c_parser_gimple_unary_expression): Likewise.
(c_parser_gimple_if_stmt): Likewise.
(c_parser_gimple_switch_stmt): Likewise.
(c_parser_gimple_return_stmt): Likewise.
(c_parser_parse_ssa_name): When name lookup fails do not build
an SSA name.  Use undeclared rather than not declared in error
reporting.

* gcc.dg/gimplefe-error-1.c: New testcase.
* gcc.dg/gimplefe-error-2.c: New testcase.
* gcc.dg/gimplefe-error-3.c: New testcase.

Co-Authored-By: Richard Biener <rguenther@suse.de>
From-SVN: r245326

re PR tree-optimization/79411 (ICE: SSA corruption (fail_abnormal_edge_coalesce))

PR tree-optimization/79411
* tree-ssa-reassoc.c (is_reassociable_op): Return false if
stmt operands are SSA_NAMEs used in abnormal phis.
(can_reassociate_p): Return false if op is SSA_NAME used in abnormal
phis.

* gcc.c-torture/compile/pr79411.c: New test.

From-SVN: r245324

Daily bump.

From-SVN: r245320

re PR c++/79143 ([new inheriting constructors] inheriting constructor fails with brace initialization)

PR c++/79143
* pt.c (instantiate_class_template_1): Copy CLASSTYPE_NON_AGGREGATE
from pattern to type.

* g++.dg/cpp1z/pr79143.C: New test.

Co-Authored-By: Jason Merrill <jason@redhat.com>
From-SVN: r245315

PR c++/79316 - default argument in deduction guide

PR c++/79350 - explicit deduction guide
* parser.c (cp_parser_constructor_declarator_p)
(cp_parser_direct_declarator): Parse deduction guides more like
constructors.
* cp-tree.h (enum special_function_kind): Add sfk_deduction_guide.
* tree.c (special_function_p): Return it.
* decl.c (check_special_function_return_type): Handle it.
(grokdeclarator, grokfndecl): Adjust.
(cp_finish_decl): Pass flags to do_auto_deduction.
* error.c (dump_decl_name): Use TFF_UNQUALIFIED_NAME.
* pt.c (dguide_name_p): Take a const_tree.
(do_class_deduction): Handle explicit.
(do_auto_deduction): Pass flags through.
(build_deduction_guide): Copy explicit flag.

From-SVN: r245314

re PR ipa/70795 (gcc/libjava/interpret.cc:1948:1: ICE: in binds_to_current_def_p, at symtab.c:2232)

PR ipa/70795
* cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
flag if needed.

From-SVN: r245312

loop-unswitch-2.c: New testcase.

* gcc.dg/loop-unswitch-2.c: New testcase.
* gcc.dg/loop-unswitch-1.c: New testcase.
* tree-ssa-loop-unswitch.c (hoist_guard): Update profile.

From-SVN: r245311

re PR c/79428 (ICE in c_parser_consume_token, at c/c-parser.c:770)

PR c/79428
* c-parser.c (c_parser_omp_ordered): Call c_parser_skip_to_pragma_eol
instead of c_parser_skip_until_found.

* c-c++-common/cilk-plus/CK/pr79428-4.c: New test.
* c-c++-common/cilk-plus/CK/pr79428-7.c: New test.
* c-c++-common/goacc/pr79428-1.c: New test.
* c-c++-common/gomp/pr79428-2.c: New test.
* c-c++-common/gomp/pr79428-5.c: New test.
* c-c++-common/gomp/pr79428-6.c: New test.
* c-c++-common/pr79428-3.c: New test.

From-SVN: r245309

S/390: Fix glob pattern for vector tests.

2017-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>

* gcc.target/s390/s390.exp: Fix test glob pattern for vector
testcases.

From-SVN: r245307

omp-offload.c (oacc_loop_auto_partitions): Use || instead of | to avoid warning.

* omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
to avoid warning.

From-SVN: r245306

re PR c/79413 (ICE in make_ssa_name_fn, at tree-ssanames.c:265)

PR c/79413
* gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
not arbitrary TREE_CONSTANT.

* gcc.c-torture/compile/pr79413.c: New test.

From-SVN: r245304

re PR c++/79429 (ICE in add_stmt, at cp/semantics.c:385)

PR c++/79429
* parser.c (cp_parser_omp_ordered): Don't check for non-pragma_stmt
non-pragma_compound context here.
(cp_parser_omp_target): Likewise.
(cp_parser_pragma): Don't call push_omp_privatization_clauses and
parsing for ordered and target omp pragmas in non-pragma_stmt
non-pragma_compound contexts.

* c-c++-common/gomp/pr79429.c: New test.
* g++.dg/gomp/pr79429.C: New test.

From-SVN: r245303

re PR c/79431 (ICE in get, at cgraph.h:397)

PR c/79431
* gimplify.c (gimplify_adjust_omp_clauses): Ignore
"omp declare target link" attribute unless is_global_var.
* omp-offload.c (find_link_var_op): Likewise.
c/
* c-parser.c (c_parser_omp_declare_target): Don't invoke
symtab_node::get on automatic variables.
cp/
* parser.c (cp_parser_oacc_declare): Formatting fix.
(cp_parser_omp_declare_target): Don't invoke symtab_node::get on
automatic variables.
testsuite/
* c-c++-common/gomp/pr79431.c: New test.

From-SVN: r245302

loop-dim-default.c: Move dg-skip-if directive into a comment.

* testsuite/libgomp.oacc-c-c++-common/loop-dim-default.c: Move
dg-skip-if directive into a comment.

From-SVN: r245301

gimplify.c (gimplify_scan_omp_clauses): No special handling for OMP_CLAUSE_TILE.

2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
    Cesar Philippidis  <cesar@codesourcery.com>
    Joseph Myers  <joseph@codesourcery.com>
    Chung-Lin Tang  <cltang@codesourcery.com>

gcc/
* gimplify.c (gimplify_scan_omp_clauses): No special handling for
OMP_CLAUSE_TILE.
(gimplify_adjust_omp_clauses): Don't delete TILE.
(gimplify_omp_for): Deal with TILE.
* internal-fn.c (expand_GOACC_TILE): New function.
* internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
(GOACC_TILE): New.
* omp-expand.c (struct oacc_collapse): Add tile and outer fields.
(expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
element fields.
(expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
avoid DIV for outermost collapse var.
(expand_oacc_for): Insert tile element loop as needed.  Adjust.
Remove out of date comments, fix whitespace.
* omp-general.c (omp_extract_for_data): Deal with tiling.
* omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
adjust OLF_DIM_BASE value.
(struct omp_for_data): Add tiling field.
* omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
(lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
for auto loops.  Remove default auto determining, moved to
oacc_loop_fixed_partitions.
* omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
stmts, add e_mask field.
(oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
(oacc_thread_numbers): Use oacc_dim_call.
(oacc_xform_tile): New.
(new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
(finish_oacc_loop): Adjust for ifns vector.
(oacc_loop_discover_walk): Append loop abstraction sites to list,
add case for GOACC_TILE fns.
(oacc_loop_xform_loop): Delete.
(oacc_loop_process): Iterate over call list directly, and add
handling for GOACC_TILE fns.
(oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
dump partitioning.
(oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
vector partitioning to outer loops.  Assign 2 partitions to loops
when available. Add TILE handling.
(oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
(execite_oacc_device_lower): Process GOACC_TILE fns, ignore unknown specs.
* tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
* tree.c (omp_clause_num_ops): Adjust TILE ops.
* tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.

gcc/c/
* c-parser.c (c_parser_omp_clause_collapse): Disallow tile.
(c_parser_oacc_clause_tile): Disallow collapse. Fix parsing and
semantic checking.
* c-parser.c (c_parser_omp_for_loop): Accept tiling constructs.

gcc/cp/
* parser.c (cp_parser_oacc_clause_tile): Disallow collapse.  Fix
parsing.  Parse constant expression. Remove semantic checking.
(cp_parser_omp_clause_collapse): Disallow tile.
(cp_parser_omp_for_loop): Deal with tile clause.  Don't emit a parse
error about missing for after already emitting one.  Use more
conventional for idiom for unbounded loop.
* pt.c (tsubst_omp_clauses): Handle OMP_CLAUSE_TILE.
* semantics.c (finish_omp_clauses): Correct TILE semantic check.
(finish_omp_for): Deal with tile clause.

gcc/fortran/
* openmp.c (resolve_omp_clauses): Error on directives
containing both tile and collapse clauses.
(resolve_oacc_loop_blocks): Represent '*' tile arguments as zero.
* trans-openmp.c (gfc_trans_omp_do): Lower tiled loops like
collapsed loops.

gcc/testsuite/
* c-c++-common/goacc/combined-directives.c: Remove xfail.
* c-c++-common/goacc/loop-auto-1.c: Adjust and add additional case.
* c-c++-common/goacc/loop-auto-2.c: New.
* c-c++-common/goacc/tile.c: Include stdbool, fix expected errors.
* c-c++-common/goacc/tile-2.c: New.
* g++.dg/goacc/template.C: Test tile subst.  Adjust erroneous uses.
* g++.dg/goacc/tile-1.C: New, check tile subst.
* gcc.dg/goacc/loop-processing-1.c: Adjust dg-final pattern.
* gfortran.dg/goacc/combined-directives.f90: Remove xfail.
* gfortran.dg/goacc/tile-1.f90: New test.
* gfortran.dg/goacc/tile-2.f90: New test.
* gfortran.dg/goacc/tile-lowering.f95: New test.

libgomp/
* testsuite/libgomp.oacc-c-c++-common/tile-1.c: New.
* testsuite/libgomp.oacc-c-c++-common/loop-auto-1.c: Adjust and
add additional case.
* testsuite/libgomp.oacc-c-c++-common/vprop.c: XFAIL under
"openacc_nvidia_accel_selected".
* libgomp.oacc-fortran/nested-function-1.f90 (test2):
Add num_workers(8) clause.

From-SVN: r245300

configure.ac (ACX_BUGURL): Update.

* configure.ac (ACX_BUGURL): Update.
* configure: Regenerate.

From-SVN: r245299

* Makefile.in (po/$(PACKAGE).pot): Adjust bug reporting URL.

From-SVN: r245298

* comperr.adb: Update FSF bug reporting URL.

From-SVN: r245297

configure.ac (ACX_BUGURL): Update.

* configure.ac (ACX_BUGURL): Update.
* configure: Regenerate.

From-SVN: r245296

re PR tree-optimization/69823 (internal compiler error: in create_pw_aff_from_tree, at graphite-sese-to-poly.c:445)

2017-02-09  Richard Biener  <rguenther@suse.de>

PR tree-optimization/69823
* graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.

* gcc.dg/graphite/pr69823.c: New testcase.

From-SVN: r245295

arc/gcc: Better creation of __NPS400__ define

The __NPS400__ define is currently created in CPP_SPEC unlike the other
target defines, which are created in arc-c.def.  Further, the current
__NPS400__ define is (currently) only created when -mcpu=nps400 is
passed, which is fine, except that if GCC is configured using
--with-cpu=nps400 then the -mcpu option is not required and the
__NPS400__ define will not be created.

This commit moves the __NPS400__ define into arc-c.def inline with all
of the other target defines, and removes the code in CPP_SPEC that used
to create the define.

In order to support the creation of the define in arc-c.def, a new
TARGET_NPS400 macro is created in arc.h.

gcc/ChangeLog:

* config/arc/arc-c.def: Add __NPS400__ definition.
* config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
(TARGET_NPS400): Define.

From-SVN: r245294

gcc/arc: Make arc_selected_cpu global

Currently we only make the base_architecture globally available, this
means we can tell if we have selected arc700/archs/etc but it's not
possible to tell if the user has selected a specific cpu variant, for
example nps400.

One problem this causes is, for example, in arc-c.def, if we want to add
an __NPS400__ define then we need a flag we can check to determine if
this is the right thing to do.

In this commit the arc_selected_cpu variable (previously local within
arc.c) has been made global.  Two other variables arc_base_cpu and
arc_selected_arch have been deleted, all of this information can be
found within (or through) arc_selected_cpu.

All uses of arc_base_cpu and arc_selected_arch have been updated.  This
commit does not introduce any new defines (like __NPS400__), this is
just a restructuring commit.

The declaration of arc_selected_cpu has moved into arc-arch.h, in
contrast to the declaration of arc_base_cpu which was previously in
arc.h.  This avoids a compilation issue when building libgcc, as the
structure and enums declared in arc-arch.h are not included for libgcc
then declaring an arc_selected_cpu (a struct type) in arc.h would result
in an unknown struct error.  We got away with this for arc_base_cpu as
that was an enum type.  The declaration of arc_selected_cpu in
arc.h could have been wrapped in a '#ifndef IN_LIBGCC2 ... #endif', but
it felt neater to simply move the declaration into arc-arch.h.

gcc/ChangeLog:

* config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
file.
(arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
pointer, arch_info.
(arc_cpu_types): Fill the arch_info field with a pointer into the
arc_arch_types table.
(arc_selected_cpu): Declare.
* config/arc/arc.c (arc_selected_cpu): Make global.
(arc_selected_arch): Delete.
(arc_base_cpu): Delete.
(arc_override_options): Remove references to deleted variables,
update access to arch information.
(ARC_OPT): Update access to arch information.
(ARC_OPTX): Likewise.
* config/arc/arc.h (arc_base_cpu): Remove declaration.
(TARGET_ARC600): Update access to arch information.
(TARGET_ARC601): Likewise.
(TARGET_ARC700): Likewise.
(TARGET_EM): Likewise.
(TARGET_HS): Likewise.
* config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
information.

From-SVN: r245293

loop-dim-default.c: Skip on hppa*-*-*.

* testsuite/libgomp.oacc-c-c++-common/loop-dim-default.c: Skip on
hppa*-*-*.
* testsuite/libgomp.oacc-c-c++-common/reduction-cplx-dbl.c: Don't
include complex.h on hppa*-*-hpux*.
* testsuite/libgomp.oacc-c-c++-common/reduction-cplx-flt.c: Likewise.

From-SVN: r245292

Daily bump.

From-SVN: r245291

* src/c++11/snprintf_lite.cc (__err): Use https for bug reporting.

From-SVN: r245288

re PR target/78604 (test case gcc.target/powerpc/p8vector-vectorize-1.c fails starting with r242750)

PR target/78604
* config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
condition/operands for integer GE/LE/GEU/LEU operations.

* gcc.target/powerpc/pr78604.c: New.

From-SVN: r245285

re PR testsuite/68972 (g++.dg/cpp1y/vla-initlist1.C test case fails on powerpc64le)

gcc/testsuite/ChangeLog:

2017-02-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>

PR target/68972
* g++.dg/cpp1y/vla-initlist1.C: Add dg-skip-if directive to
disable this test on power architecture.

From-SVN: r245280

Update links and references in libstdc++ manual

* doc/xml/manual/policy_data_structures.xml: Fix spelling of author's
name.
* doc/xml/manual/policy_data_structures_biblio.xml: Likewise. Remove
broken links to texts that are no longer online.
* doc/xml/manual/profile_mode.xml: Update links to CGO 2009 paper and
LCPC 2006 paper.
* doc/xml/manual/using.xml: Update links to memory model information.
* doc/xml/manual/using_exceptions.xml: Update link to "Appendix E:
Standard-Library Exception Safety".
* doc/html/*: Regenerate.

From-SVN: r245277

rs6000: Fix spelling of AltiVec in rs6000.opt (PR79397)

It was spelled Altivec in two user-facing messages in rs6000.opt; let's
be consistent.

PR translation/79397
* config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
of AltiVec.

From-SVN: r245276

[PR 79375] Avoid passing NULL by reference

2017-02-08  Martin Jambor  <mjambor@suse.cz>

PR ipa/79375
* ipa-prop.c (ipa_alloc_node_params): Make static, return bool
whether allocation happened.
(ipa_initialize_node_params): Do not call ipa_alloc_node_params if
nothing was allocated.

From-SVN: r245275

re PR tree-optimization/71824 (ICE when compiling libiberty with Graphite loop optimizations)

2017-02-08  Richard Biener  <rguenther@suse.de>

PR tree-optimization/71824
PR tree-optimization/79409
* gcc.dg/graphite/pr71824-3.c: New testcase.

From-SVN: r245274

re PR tree-optimization/79408 (Missed VRP optimization of integer modulo)

PR tree-optimization/79408
* tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
constant, but SSA_NAME with a known integer range, use the minimum
of that range instead of op1 to determine if modulo can be replaced
with its first operand.

* gcc.dg/tree-ssa/pr79408.c: New test.

From-SVN: r245273

[riscv] Fix build due to INT16_MAX issue

* config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.

From-SVN: r245272

re PR tree-optimization/71824 (ICE when compiling libiberty with Graphite loop optimizations)

2017-02-08  Richard Biener  <rguenther@suse.de>

PR tree-optimization/71824
* graphite-scop-detection.c (scop_detection::build_scop_breadth):
Check all loops contained in the merged region.

* gcc.dg/graphite/pr71824-2.c: New testcase.

From-SVN: r245270

profile_mode.xml: Unbreak link to "Optimizing Sorting with Machine Learning Algorithms".

        * doc/xml/manual/profile_mode.xml: Unbreak link to
        "Optimizing Sorting with Machine Learning Algorithms".

From-SVN: r245269

* src/c++11/snprintf_lite.cc (__err): Update bug reporting URL.

From-SVN: r245268

aarch64.md (popcount<mode>2): New pattern.

2016-02-07  Andrew Pinski  <apinski@cavium.com>

gcc
* config/aarch64/aarch64.md (popcount<mode>2): New pattern.

gcc/testsuite
* gcc.target/aarch64/popcount.c : New Testcase.

From-SVN: r245267

aarch64-cores.def (thunderx): Disable LSE.

2017-02-07  Andrew Pinski  <apinski@cavium.com>

        * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
        (thunderxt88): Likewise.
        (thunderxt81): Disable LSE and change v8.1 to v8.
        (thunderxt83): Likewise.

From-SVN: r245266

Daily bump.

From-SVN: r245264

abi.xml: Update link to "Sun Studio 11: C++ Migration Guide".

* doc/xml/manual/abi.xml: Update link to "Sun Studio 11: C++
Migration Guide".

From-SVN: r245261

* update_web_docs_svn (MANUALS): Add libitm.

From-SVN: r245260

lwg-active.html: Remove.

* doc/html/ext/lwg-active.html: Remove.
* doc/html/ext/lwg-closed.html: Ditto.
* doc/html/ext/lwg-defects.html: Ditto.

* doc/Makefile.am (xml_extradir): Remove.
(xml_extra): Ditto.
(stamp-html-docbook-lwg): Remove recipe...
(stamp-html-docbook-data): ...and its use here.
* doc/Makefile.in: Regenerate.

* doc/xml/manual/intro.xml: Shorten two paragraphs explaining
the relationship to the upstream working group.
Replace a local link to ../ext/lwg-active.html by the upstream one.
Replace all reference to ../ext/lwg-defects.html by a new entity
&DR; which refers to the upstream address.

From-SVN: r245258

* doc/xml/manual/status_cxx2017.xml: Fix link to N4284.

From-SVN: r245257

re PR middle-end/79399 (GCC fails to compile big source at -O0)

PR middle-end/79399
* ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
type from int to size_t.
* ira-costs.c (struct_costs_size): Change type from int to size_t.

Co-Authored-By: Richard Biener <rguenther@suse.de>
From-SVN: r245256

trans-types.c (gfc_get_int_kind_from_width_isofortranen): Choose REAL type with the widest precision if...

2017-02-04  Steven G. Kargl  <kargl@gcc.gnu.org>

* trans-types.c (gfc_get_int_kind_from_width_isofortranen):  Choose
REAL type with the widest precision if two (or more) have the same
storage size.

From-SVN: r245255

* MAINTAINERS: Adjust bug reporting URL.

From-SVN: r245253

re PR rtl-optimization/79386 (ICE: segmentation fault in cprop w/ -O2 on 32-bit BE powerpc)

PR rtl-optimization/79386
* cprop.c (bypass_conditional_jumps): Initialize
bypass_last_basic_block already before splitting bbs after
unconditional traps...
(bypass_conditional_jumps): ... rather than here.

* gcc.c-torture/compile/pr79386.c: New test.

From-SVN: r245251

method.c (synthesized_method_base_walk): New.

* method.c (synthesized_method_base_walk): New.  Broken out of ...
(synthesized_method_walk): ... here.  Call it.  Cleanup
initializations.

From-SVN: r245250

Only xfail g++.dg/tls/thread_local-order2.C on Solaris

2017-02-07  Dominik Vogt  <vogt@linux.vnet.ibm.com>
    Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* g++.dg/tls/thread_local-order2.C: Only xfail execution on
*-*-solaris*.

Co-Authored-By: Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
From-SVN: r245249

re PR target/79299 (Operand size mismatch for `vpgatherqd' w/ -O3 -masm=intel -mavx512bw)

PR target/79299
* config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
(*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
*avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
fix -masm=intel patterns.

* gcc.target/i386/avx512vl-pr79299-1.c: New test.
* gcc.target/i386/avx512vl-pr79299-2.c: New test.

From-SVN: r245248

re PR tree-optimization/79256 (FAIL: gcc.dg/vect/pr25413a.c execution test)

2017-02-07  Richard Biener  <rguenther@suse.de>

PR tree-optimization/79256
PR middle-end/79278
* builtins.c (get_object_alignment_2): Use min_align_of_type
to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
and ADJUST_FIELD_ALIGN.

* doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
type parameter.
* doc/tm.texi: Regenerate.
* stor-layout.c (layout_decl): Adjust.
(update_alignment_for_field): Likewise.
(place_field): Likewise.
(min_align_of_type): Likewise.
* config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
* config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
* config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
* config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
* config/frv/frv.c (frv_adjust_field_align): Likewise.
* config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
* config/i386/i386.c (x86_field_alignment): Likewise.
* config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
* config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
* config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
* config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
* config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
* config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
 Likewise.

go/
* go-backend.c (go_field_alignment): Adjust.

libobjc/
* encoding.c (objc_layout_structure_next_member): Adjust
ADJUST_FIELD_ALIGN usage.

Revert
2017-01-30  Richard Biener  <rguenther@suse.de>

PR tree-optimization/79256
* targhooks.c (default_builtin_vector_alignment_reachable): Honor
BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
alignment on TYPE.

From-SVN: r245245

gimple-parser.c (c_parser_gimple_expr_list): Simplify.

2017-02-07  Richard Biener  <rguenther@suse.de>

c/
* gimple-parser.c (c_parser_gimple_expr_list): Simplify.
(c_parser_gimple_postfix_expression_after_primary):
Do not use c_build_function_call_vec to avoid folding and promotion.
Simplify.

* gcc.dg/gimplefe-23.c: New testcase.
* gcc.dg/gimplefe-24.c: Likewise.

From-SVN: r245244

MIPS: Fix mode mismatch error between Loongson builtin arguments and insn
operands.

gcc/

* config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
builtins to SImode and emit a zero-extend, if necessary.

From-SVN: r245243

[AArch64,testsuite] Fix epilogue scan after ifcombine update.

2017-02-07  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/test_frame_1.c: Scan epilogue with
scan-assembler instead of scan-assembler-times.
* gcc.target/aarch64/test_frame_10.c: Likewise.
* gcc.target/aarch64/test_frame_12.c: Likewise.
* gcc.target/aarch64/test_frame_2.c: Likewise.
* gcc.target/aarch64/test_frame_4.c: Likewise.
* gcc.target/aarch64/test_frame_6.c: Likewise.
* gcc.target/aarch64/test_frame_7.c: Likewise.
* gcc.target/aarch64/test_frame_8.c: Likewise.

From-SVN: r245242

Require ifunc support in gcc.target/i386/pr78419.c

* gcc.target/i386/pr78419.c: Require ifunc support.

From-SVN: r245241

* gcc.dg/torture/stackalign/builtin-apply-2.c: Fix skip-if syntax.

From-SVN: r245240

re PR c++/79360 (ICE with NSDMI and enum in union)

Fix PR c++/79360

gcc/cp/ChangeLog:

PR c++/79360
* typeck2.c (process_init_constructor_union): Consider only
FIELD_DECLs when looking for an NSDMI.

gcc/testsuite/ChangeLog:

PR c++/79360
* g++.dg/cpp1y/nsdmi-union2.C: New test.

From-SVN: r245239

Daily bump.

From-SVN: r245237

* de.po, fr.po, sv.po: Update.

From-SVN: r245233

Alphebetize RISC-V Options section

2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>

        * docs/invoke.texi (RISC-V Options): Alphabetize.

From-SVN: r245231

Use two spaces to separate options in the RISC-V docs

gcc/ChangeLog:

2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>

        * doc/invoke.texi (RISC-V Options): Use two spaces to separate
        options.

From-SVN: r245230

RISC-V Port: contrib

2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>

        * config-list.mki (LIST): Add riscv32-unknown-linux-gun and
        riscv64-unknown-linux-gnu.

From-SVN: r245229

RISC-V Port: gcc/testsuite

gcc/testsuite/ChangeLog:

2017-02-06  Kito Cheng  <kito.cheng@gmail.com>

        * lib/target-supports.exp: Define the RISC-V target.
        * g++.dg/cpp0x/constexpr-rom.C: Skip on RISC-V
        * gcc.dg/builtin-apply2.c: Likewise.
        * gcc.dg/ifcvt-4.c: Likewise.
        * gcc.dg/loop-8.c: Likewise.
        * gcc.dg/sibcall-10.c: Likewise.
        * gcc.dg/sibcall-9.c: Likewise.
        * gcc.dg/torture/stackalign/builtin-apply-2.c: Likewise.
        * gcc.dg/tree-ssa/20040204-1.c: Likewise.
        * gcc.dg/tree-ssa/ssa-dom-cse-2.c: Likewise.
        * gcc.dg/tree-ssa/ssa-fre-3.c: Likewise.
        * gcc.c-torture/execute/20101011-1.c: Define DO_TEST on RISC-V.
        * gcc.dg/20020312-2.c: Don't define PIC_REG on RISC-V.
        * gcc.dg/stack-usage-1.c: Define SIZE on RISC-V.

From-SVN: r245228

RISC-V Port: libatomic

2017-02-06  Palmer Dabbelt <palmer@dabbelt.com>

        * configure.tgt: Add RISC-V tuple.

From-SVN: r245227

RISC-V Port: libgcc

libgcc/ChangeLog:

2017-02-06  Palmer Dabbelt <palmer@dabbelt.com>

        * config.host: Add RISC-V tuples.
        * config/riscv/atomic.c: New file.
        * config/riscv/crti.S: Likewise.
        * config/riscv/crtn.S: Likewise.
        * config/riscv/div.S: Likewise.
        * config/riscv/linux-unwind.h: Likewise.
        * config/riscv/muldi3.S: Likewise.
        * config/riscv/multi3.S: Likewise.
        * config/riscv/save-restore.S: Likewise.
        * config/riscv/sfp-machine.h: Likewise.
        * config/riscv/t-elf: Likewise.
        * config/riscv/t-elf32: Likewise.
        * config/riscv/t-elf64: Likewise.
        * config/riscv/t-softfp32: Likewise.
        * config/riscv/t-softfp64: Likewise.

From-SVN: r245226

RISC-V Port: Regenerate gcc/configure

From-SVN: r245225

RISC-V Port: gcc

gcc/ChangeLog:

2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>

        * config/riscv/riscv.c: New file.
        * gcc/common/config/riscv/riscv-common.c: Likewise.
        * config.gcc: Likewise.
        * config/riscv/constraints.md: Likewise.
        * config/riscv/elf.h: Likewise.
        * config/riscv/generic.md: Likewise.
        * config/riscv/linux.h: Likewise.
        * config/riscv/multilib-generator: Likewise.
        * config/riscv/peephole.md: Likewise.
        * config/riscv/pic.md: Likewise.
        * config/riscv/predicates.md: Likewise.
        * config/riscv/riscv-builtins.c: Likewise.
        * config/riscv/riscv-c.c: Likewise.
        * config/riscv/riscv-ftypes.def: Likewise.
        * config/riscv/riscv-modes.def: Likewise.
        * config/riscv/riscv-opts.h: Likewise.
        * config/riscv/riscv-protos.h: Likewise.
        * config/riscv/riscv.h: Likewise.
        * config/riscv/riscv.md: Likewise.
        * config/riscv/riscv.opt: Likewise.
        * config/riscv/sync.md: Likewise.
        * config/riscv/t-elf-multilib: Likewise.
        * config/riscv/t-linux: Likewise.
        * config/riscv/t-linux-multilib: Likewise.
        * config/riscv/t-riscv: Likewise.
        * configure.ac: Likewise.
        * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
        Waterman as RISC-V maintainers.
        * doc/install.texi: Add RISC-V entries.
        * doc/invoke.texi: Add RISC-V options section.
        * doc/md.texi: Add RISC-V constraints section.

From-SVN: r245224

PR c++/71193 - incomplete types in templates

* parser.c (cp_parser_postfix_dot_deref_expression): In a template
handle incomplete type by pedwarning and then treating as dependent.

From-SVN: r245223

re PR target/66144 (vector element operator produces very bad code)

[gcc]
2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/66144
* config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
false values to be constant vectors with all 0 or all 1 bits set.
(vcondu<mode><mode>): Likewise.
* config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
predicate.
(fpmask_comparison_operator): Update comment.
(vecint_comparison_operator): New predicate.
* config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
vector conditionals when the true and false values are constant
vectors with all 0 bits or all 1 bits set.

[gcc/testsuite]
2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/66144
* gcc.target/powerpc/pr66144-1.c: New test.
* gcc.target/powerpc/pr66144-2.c: Likewise.
* gcc.target/powerpc/pr66144-3.c: Likewise.

From-SVN: r245222

PR tree-optimization/ 79376 - wrong lower bound with %s and non-constant
strings in -Wformat-overflow

gcc/ChangeLog:
* gimple-fold.c (get_range_strlen): Set the minimum length to zero.

From-SVN: r245221

re PR c++/79379 (ICE with #pragma GCC ivdep)

PR c++/79379
* constexpr.c (cxx_eval_constant_expression): Handle ANNOTATE_EXPR.
(potential_constant_expression_1): Likewise.

* g++.dg/cpp1y/constexpr-79379.C: New test.

From-SVN: r245220

re PR c++/79377 (ICE with increment operator in -fpermissive mode)

PR c++/79377
* tree.c (build_min_non_dep_op_overload): For POST{INC,DEC}REMENT_EXPR
allow one fewer than expected arguments if flag_permissive.

* g++.dg/lookup/pr79377.C: New test.

From-SVN: r245219

re PR c++/79372 (ICE on C++ code with illegal decomposition declaration on x86_64-linux-gnu: in tsubst_decomp_names, at cp/pt.c:15599)

PR c++/79372
* decl.c (cp_finish_decomp): On error set decl type to error_mark_node.
* pt.c (tsubst_expr): Don't call tsubst_decomp_names on decompositions
with error_mark_node type.

* g++.dg/cpp1z/decomp25.C: New test.

From-SVN: r245218

sse.md (vector modes -> vec_extract* splitter): Use explicit subreg RTX with operand 1.

* config/i386/sse.md (vector modes -> vec_extract* splitter): Use
explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
to simplify split condition.

From-SVN: r245217

omp-expand.c (oxpand_omp_atomic_fetch_op, [...]): Return false if can_atomic_load_p is false.

* omp-expand.c (oxpand_omp_atomic_fetch_op,
expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
false.

From-SVN: r245216

sched: Do not move expensive insns speculatively (PR68664)

Scheduling should never move very expensive instructions to places they
are executed more frequently.  This patch fixes that, reducing the
execution time of c-ray by over 40% (I tested on a BE Power7 system).

This introduces a new target hook sched.can_speculate_insn which returns
whether the scheduler is allowed to speculate a given instruction.  The
rs6000 implementation disallows all divide and square root instructions.

PR rtl-optimization/68664
* target.def (can_speculate_insn): New hook.
* doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
* doc/tm.texi: Regenerate.
* sched-rgn.c (can_schedule_ready_p): Use the new hook.
* config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
(rs6000_sched_can_speculate_insn): New function.

From-SVN: r245215

re PR tree-optimization/79284 (ICE on valid code at -O3 on x86_64-linux-gnu: verify_gimple failed)

PR tree-optimization/79284
* tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
* tree-vect-stmts.c (vect_get_vec_def_for_operand,
vectorizable_mask_load_store, vectorizable_operation,
vect_is_simple_cond, get_same_sized_vectype): Use it instead
of comparing TREE_CODE of a type against BOOLEAN_TYPE.
* tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
* tree-vect-slp.c (vect_get_constant_vectors): Likewise.
* tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
is_gimple_assign (stmt).  Replace another such test with
is_gimple_assign (stmt).
testsuite/
* gcc.c-torture/compile/pr79284.c: New test.

From-SVN: r245214

re PR c++/70448 (ICE on invalid code on x86_64-linux-gnu: unexpected expression ‘f’ of kind overload)

2017-02-06  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/70448
* g++.dg/cpp0x/pr70448.C: New.

From-SVN: r245213

Fix test failure at -O0 by pruning output

PR libstdc++/79323
* testsuite/20_util/duration/literals/range.cc: Prune extra output
at -O0.

From-SVN: r245211

PR 78348: Limit amount of unrolling in test.

gcc/testsuite/ChangeLog:

2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>

PR 78348
* gcc.dg/tree-ssa/ldist-24.c: Add "--param max-completely-peel-times=8"
to dg-options.

From-SVN: r245210

re PR rtl-optimization/78883 ([avr] ICE triggered by change to combine.c (r243578))

gcc/
PR target/78883
* config/avr/avr.c (rtl-iter.h): Include it.
(TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
(avr_legitimate_combined_insn): ...and implementation.
gcc/testsuite/
PR target/78883
* gcc.c-torture/compile/pr78883.c: New test.

From-SVN: r245209

S/390: Use macros from hwint.h where possible.

gcc/ChangeLog:

2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>

* config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
* config/s390/s390.c (s390_const_operand_ok)
(s390_canonicalize_comparison, s390_extract_part)
(s390_single_part, s390_contiguous_bitmask_nowrap_p)
(s390_contiguous_bitmask_p, s390_rtx_costs)
(legitimize_pic_address): Likewise.
* config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
* config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
* config/s390/vector.md ("vec_vfenes<mode>"): Likewise.

From-SVN: r245207