Avoid unnecessarily numbering cloned symbols.

2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>

* cgraph.h (clone_function_name_1): Replaced by new
  clone_function_name_numbered that takes name as string; for
  privatize_symbol_name_1 use only.
  (clone_function_name): Renamed to
  clone_function_name_numbered to be explicit about numbering.
  (clone_function_name): New two-argument function that does
  not number its output.
  (clone_function_name): New three-argument function that
  takes a number to append to its output.
* cgraphclones.c (duplicate_thunk_for_node):
  (clone_function_name_1): Renamed.
  (clone_function_name_numbered): Two new functions.
  (clone_function_name): Improved documentation.
  (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
* config/rs6000/rs6000.c (make_resolver_func): Ditto.
* final.c (final_scan_insn_1): Use the new clone_function_name
  without numbering.
* multiple_target.c (create_dispatcher_calls): Ditto.
  (create_target_clone): Ditto.
* omp-expand.c (grid_expand_target_grid_body): Ditto.
* omp-low.c (create_omp_child_function_name): Ditto.
* omp-simd-clone.c (simd_clone_create): Ditto.
* symtab.c (simd_symtab_node::noninterposable_alias): Use the
  new clone_function_name without numbering.
2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>

* lto-partition.c (privatize_symbol_name_1): Use
  clone_function_name_numbered.
2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>

* gcc.dg/tree-prof/cold_partition_label.c: Update for cold
  section names without numbers.
* gcc.dg/tree-prof/section-attr-1.c: Ditto.
* gcc.dg/tree-prof/section-attr-2.c: Ditto.
* gcc.dg/tree-prof/section-attr-3.c: Ditto.

From-SVN: r265621

Don't allow the pool allocator to be configured to allocate zero-sized objects

PR bootstrap/87747 would have been significantly easier to track down if
the pool allocator had faulted an attempt to configure it to allocate
zero-sized objects.  Instead, this slipped through and we later hit
memory corruption when the assumed size turned out to be different to
the configured size.

While, theoretically, there might be a use case for this, it seems
unlikely to me that GCC would have such a use.  So this patch adds a
checking assert that the object size is not zero.

* alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
Assert that the allocation size is not zero.

From-SVN: r265620

re PR tree-optimization/87800 (CPU2006 416.gamess failed to build with LTO)

2018-10-30  Richard Biener  <rguenther@suse.de>

PR tree-optimization/87800
* tree-vect-slp.c (vect_build_slp_tree_2): Reject any
non-induction or reduction PHIs.

From-SVN: r265619

Fixes bug 87330 by invoking  df_note_add_problem to recompute REG_DEAD and REG_UNUSED notes before analysis.

From-SVN: r265618

libgcc: properly destroy mutexes on VxWorks

Just as one needs run-time initialization of mutexes, one needs to
destroy them properly to allow the OS to release resources associated
with the semaphore.

From-SVN: r265616

Fix typo in documentation of __builtin_expect_with_probability.

2018-10-30  Martin Liska  <mliska@suse.cz>

* doc/extend.texi: Fix typo in documentation
of __builtin_expect_with_probability.

From-SVN: r265615

Daily bump.

From-SVN: r265614

Fix ICE in get_substring_ranges_for_loc on __FILE__ (PR c++/87721)

PR c++/87721 reports a crash in get_substring_ranges_for_loc introduced
by r265271, my fix for PR 87562.

The new issue occurs when attempting to get a location with a string
literal inside a macro in which the first token is __FILE__ (formed via
concatenation).  Attempting to get the spelling location of __FILE__
fails, leading to NULL for start_ord_map and final_ord_map, and thus
a NULL pointer dereference.

Given that our "on-demand" substring locations approach reparses the
string literals, there isn't a good way to access the locations inside
such string literals: attempting to reparse __FILE__ fails with a
"missing open quote".

This patch applies the easy fix by gracefully rejecting the case where
the spelling locations for the start or finish give us NULL maps.

gcc/ChangeLog:
PR c++/87721
* input.c (get_substring_ranges_for_loc): Detect if
linemap_resolve_location gives us a NULL map, and reject
this case.

gcc/testsuite/ChangeLog:
PR c++/87721
* c-c++-common/substring-location-PR-87721.c: New test.
* gcc.dg/plugin/diagnostic-test-string-literals-1.c: Add test for
PR 87721.
* gcc.dg/plugin/diagnostic_plugin_test_string_literals.c
(test_string_literals): Fold the index arguments before checking
for INTEGER_CST.

From-SVN: r265611

C++: simplify output from suggest_alternatives_for

In the C++ FE, after emitting various errors about unrecognized names,
the parser can call
  suggest_alternatives_for
and/or
  suggest_alternative_in_explicit_scope.
These can issue zero or more suggestions for the unrecognized name,
or various other "note" diagnostics suggesting how to fix the problem.

For example, currently g++ emits:

t.cc:12:3: error: 'gtk_widget_showall' was not declared in this scope
12 |   gtk_widget_showall (w);
   |   ^~~~~~~~~~~~~~~~~~
t.cc:12:3: note: suggested alternative: 'gtk_widget_show_all'
12 |   gtk_widget_showall (w);
   |   ^~~~~~~~~~~~~~~~~~
   |   gtk_widget_show_all

This patch consolidates the common case when there is a single
candidate, so that the error can issue a fix-it hint directly.

This simplifies the above to:

t.cc:12:3: error: 'gtk_widget_showall' was not declared in this scope;
 did you mean 'gtk_widget_show_all'?
12 |   gtk_widget_showall (w);
   |   ^~~~~~~~~~~~~~~~~~
   |   gtk_widget_show_all

omitting the second "note" diagnostic.

Doing so requires changing the above "suggest_" functions so that
rather than being called after "error" and emitting a note directly,
they are called before the "error", and return a name_hint, which
can contain a suggestion and/or a deferred diagnostic.  The "single
candidate" case is handled via a suggestion, and the "multiple
candidates" case via a new subclass of deferred_diagnostic.

There was some complication due to the fact that we don't always have
enough location information to issue a fix-it hint.  Specifically,
for the case in qualified_name_lookup_error, the location is that of
the name, but the location of the qualifier prefix isn't reliably
available.  For some hints, e.g. spell-corrections, the replacement
is of the name, and for others, e.g. parent namespaces, it's for the
qualified name.  The patch addresses this by splitting this case out
into a new "suggest_alternatives_in_other_namespaces" function, for
which fix-it hints aren't issued.

Another complication is that of emitting a note when
  --param cxx-max-namespaces-for-diagnostic-help
is reached.  The patch emulates the existing behavior by emitting
the note from a deferred_diagnostic.  This potentially needs to
co-exist with another deferred_diagnostic, so it works as a decorator
around any other such deferred_diagnostic.  Doing so requires slightly
extending class name_hint.

On adding test coverage for the various cases, I discovered that
after emitting a "FOO is not a namespace-name" error, we also emit
a "expected namespace-name before" error.  The patch removes this
second error for the case where it's redundant, simplifying this case
from e.g.:

spellcheck-ns.C:10:24: error: 'inner_ms' is not a namespace-name
10 | using namespace outer::inner_ms;
   |                        ^~~~~~~~
spellcheck-ns.C:10:24: note: suggested alternative: 'inner_ns'
10 | using namespace outer::inner_ms;
   |                        ^~~~~~~~
   |                        inner_ns
spellcheck-ns.C:10:32: error: expected namespace-name before ';' token
10 | using namespace outer::inner_ms;
   |                                ^

to:

spellcheck-ns.C:10:24: error: 'inner_ms' is not a namespace-name;
 did you mean 'inner_ns'?
10 | using namespace outer::inner_ms;
   |                        ^~~~~~~~
   |                        inner_ns

include/ChangeLog:
* unique-ptr.h (gnu::move): Generalize so it applies to all
lvalue references, rather than just to unique_ptr values.

gcc/c-family/ChangeLog:
* name-hint.h (name_hint::take_deferred): New member function.

gcc/c/ChangeLog:
* c-decl.c (implicit_decl_warning): Update "is there a suggestion"
logic for change to name_hint::operator bool.
(undeclared_variable): Likewise.
* c-parser.c (c_parser_declaration_or_fndef): Likewise.
(c_parser_parameter_declaration): Likewise.

gcc/cp/ChangeLog:
* cp-name-hint.h: New file.
* cp-tree.h (expr_to_string): New decl.
(suggest_alternatives_for): Move to cp-name-hint.h, changing
return type from bool to name_hint.
(suggest_alternative_in_explicit_scope): Likewise.
* error.c: Define INCLUDE_UNIQUE_PTR.  Include "cp-name-hint.h".
(expr_to_string): Make non-static.
(qualified_name_lookup_error): For the non-"::" case, take
responsibity for issuing any suggestion from
suggest_alternative_in_explicit_scope, as it changes from
returning a bool to returning a name_hint.  Replace fallback call
to suggest_alternatives_for to a call to
suggest_alternatives_in_other_namespaces, capturing the fact that
we don't have enough location information to issue a fix-it hint
for this case.  Update the error to support emitting a fix-it hint
where appropriate.  For the "::" case, take responsibility for
issuing any suggestion from suggest_alternatives_for, supporting
emitting a fix-it hint.
* lex.c: Define INCLUDE_UNIQUE_PTR.  Include "gcc-rich-location.h"
and "cp-name-hint.h".
(unqualified_name_lookup_error): Take responsibility for issuing
any suggestion from suggest_alternatives_for, supporting emitting
a fix-it hint.
* name-lookup.c (class namespace_limit_reached): New subclass of
deferred_diagnostic.
(class show_candidate_location): Likewise.
(class suggest_alternatives): Likewise.
(class namespace_hints): New class.
(suggest_alternatives_for): Convert return type from bool to
name_hint, replacing all direct diagnostic emission by setting
suggestions on the return value, or creating deferred diagnostics.
Specifically, split out initial traversal of namespaces into
namespace_hints' ctor, and maybe_decorate_with_limit, and move the
rest of the implementation to
namespace_hints::convert_candidates_to_name_hint and
suggest_alternatives_for_1.
(namespace_hints::namespace_hints): New ctor, adapted from
suggest_alternatives_for's initial namespace traversal, storing
location and name, and converting locals "candidates", "limited"
and "limit" into members.
(namespace_hints::convert_candidates_to_name_hint): New member
function.
(namespace_hints::maybe_decorate_with_limit): New member function.
(suggest_alternatives_for_1): New function, based on second half
of old implementation of suggest_alternatives_for, converting from
immediate emission of suggestions to using name_hint.
(suggest_alternatives_in_other_namespaces): New function.
(maybe_suggest_missing_std_header): Convert from immediate
emission of suggestions to using name_hint, moving emission
implementation to...
(class missing_std_header): New subclass of deferred_diagnostic.
(maybe_suggest_missing_header): Convert return type from bool to
name_hint.
(suggest_alternative_in_explicit_scope): Convert from immediate
emission of suggestions to using name_hint.
* parser.c: Replace include of "c-family/name-hint.h" with
"cp-name-hint.h".
(cp_parser_diagnose_invalid_type_name): Update
"is there a suggestion" logic for change to
name_hint::operator bool.  Take responsibility for emitting
fix-it hints from suggest_alternative_in_explicit_scope.
(cp_parser_namespace_name): Take responsibility for emitting
fix-it hints from suggest_alternative_in_explicit_scope.  Don't
emit the "expected namespace-name" error if we've already emitted
an "is not a namespace-name" error.

gcc/testsuite/ChangeLog:
* c-c++-common/spellcheck-reserved.c: Update expected output for
C++ for merger of "did you mean" suggestions into the error
message.
* g++.dg/ext/builtin3.C: Update expected output for merger of "did
you mean" suggestion into the error.
* g++.dg/lookup/error1.C: Likewise.
* g++.dg/lookup/pr77549.C: Likewise.
* g++.dg/lookup/pr80913.C: Likewise.
* g++.dg/lookup/suggestions1.C: Likewise.
* g++.dg/lookup/suggestions2.C: New test.
* g++.dg/overload/koenig1.C: Update expected output as above.
* g++.dg/spellcheck-identifiers-2.C: Likewise.
* g++.dg/spellcheck-identifiers.C: Likewise.
* g++.dg/spellcheck-ns.C: New test.
* g++.dg/spellcheck-pr77829.C: Update expected output as above.
* g++.dg/spellcheck-pr78656.C: Likewise.
* g++.dg/spellcheck-pr79298.C: Likewise, adding
-fdiagnostics-show-caret to options.
* g++.dg/spellcheck-pr80177.C: Likewise.
* g++.dg/spellcheck-single-vs-multiple.C: New test.
* g++.dg/spellcheck-typenames.C: Update expected output as above.
* g++.dg/template/static10.C: Likewise.
* g++.old-deja/g++.mike/ns5.C: Likewise.
* g++.old-deja/g++.mike/ns7.C: Likewise.
* g++.old-deja/g++.ns/koenig5.C: Likewise.
* g++.old-deja/g++.other/lineno5.C: Likewise.

libstdc++-v3/ChangeLog:
* testsuite/17_intro/using_namespace_std_exp_neg.cc: Remove
"expected namespace-name before" error.
* testsuite/17_intro/using_namespace_std_tr1_neg.cc: Likewise.

From-SVN: r265610

Folding and check_function_arguments

This patch eliminates the arglocs array I introduced to build_over_call
in r264887, and eliminates the call to maybe_constant_value when building
"fargs" (thus retaining location wrapper nodes).

Instead, this patch requires that any checks within
check_function_arguments that need folded arguments do their own folding.

Of the various checks:
(a) check_function_nonnull already calls fold_for_warn,
(b) check_function_format doesn't need folding
(c) check_function_sentinel needs fold_for_warn in one place, which the
patch adds, and
(d) check_function_restrict needs per-argument folding, which the patch
adds.  Given that it scans before and after resetting TREE_VISITED on
each argument, it seemed best to make a copy of the array, folding each
argument from the outset, rather than repeatedly calling fold_for_warn;

gcc/c-family/ChangeLog:
PR c++/56856
* c-common.c (check_function_sentinel): Call fold_for_warn on the
argument.
(check_function_restrict): Rename param "argarray" to
"unfolded_argarray", and make a copy named "argarray", calling
fold_for_warn on each argument.
(check_function_arguments): Add note about responsibility for
folding the arguments.

gcc/cp/ChangeLog:
PR c++/56856
* call.c (build_over_call): Eliminate the "arglocs" array, and the
call to maybe_constant_value when building "fargs".

From-SVN: r265609

decl.c (create_array_type_for_decl): Add location_t parameter and use it.

/cp
2018-10-29  Paolo Carlini  <paolo.carlini@oracle.com>

* decl.c (create_array_type_for_decl): Add location_t parameter
and use it.
(grokdeclarator): Adjust call.

/testsuite
2018-10-29  Paolo Carlini  <paolo.carlini@oracle.com>

* g++.dg/cpp0x/auto24.C: Test location too.
* g++.dg/cpp0x/auto3.C: Likewise.
* g++.dg/cpp0x/auto42.C: Likewise.
* g++.dg/cpp0x/initlist57.C: Likewise.
* g++.dg/cpp0x/initlist75.C: Likewise.
* g++.dg/cpp0x/initlist80.C: Likewise.
* g++.dg/cpp0x/lambda/lambda-ice13.C: Likewise.
* g++.old-deja/g++.brendan/array-refs.C: Likewise.
* g++.old-deja/g++.bugs/900322_01.C: Likewise.
* g++.old-deja/g++.bugs/900519_07.C: Likewise.
* g++.old-deja/g++.other/typeck1.C: Likewise.

From-SVN: r265608

config.gcc (xstormy16-*-elf): Set tm_d_file.

gcc/ChangeLog:

2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>

* config.gcc (xstormy16-*-elf): Set tm_d_file.

From-SVN: r265607

ux.texi: move "Quoting" and "Fix-it hints" from DiagnosticsGuidelines wiki page

gcc/ChangeLog:
* doc/ux.texi (Quoting): New subsection, adapted from material at
https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
MartinSebor and ManuelLopezIbanez.
(Fix-it hints): Note that fix-it hints shouldn't be marked for
translation.

Co-Authored-By: Manuel López-Ibáñez <manu@gcc.gnu.org>
Co-Authored-By: Martin Sebor <msebor@redhat.com>
From-SVN: r265606

re PR c++/87469 (ice in record_estimate, at tree-ssa-loop-niter.c:3271)

gcc/testsuite/ChangeLog:

2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>

PR middle-end/87469
* g++.dg/pr87469.C: New test.

gcc/ChangeLog:

2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>

PR middle-end/87469
* tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
max value.

From-SVN: r265605

Add ChangeLog entry for hunk introduced together with another one

From-SVN: r265604

[rs6000] Remove inaccurate comment

DEBUG doesn't actually "replace abort with printf on error",
it just enables debugging output.  Just remove the comment.

[gcc/testsuite]

2018-10-29  Paul A. Clarke  <pc@us.ibm.com>

* gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.

From-SVN: r265603

* doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.

From-SVN: r265602

[rs6000] Consistently use '__vector' instead of 'vector'

Revision r265535 committed changes that used 'vector' instead of the
preferred '__vector'.  There is a reason that '__vector' is preferred,
because it ensures no conflicts with C++ namespace.  Indeed,
gcc/config/rs6000/xmmintrin.h undefines it, leading to errors:

  gcc/include/xmmintrin.h:999:20: error: 'vector' undeclared (first use in this function); did you mean 'vec_or'?
  gcc/include/xmmintrin.h:999:20: note: each undeclared identifier is reported only once for each function it appears in
  gcc/include/xmmintrin.h:999:26: error: expected ')' before 'long'
  gcc/include/xmmintrin.h:999:37: error: expected ')' before 'result'

Also fixed a few whitespace issues.

[gcc]

2018-10-29  Paul A. Clarke  <pc@us.ibm.com>

* gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
_mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
_mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
_mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
_mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
_mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
_mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
_mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
Change 'vector' to '__vector'.
* gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
_mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
_mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
_mm_avg_pu16): Likewise.  And, whitespace corrections.

From-SVN: r265601

PR c++/87594 - constexpr rejects-valid with range-based for.

* constexpr.c (potential_constant_expression_1): If the condition
can't be evaluated, return true.

* g++.dg/cpp1y/constexpr-loop8.C: New test.

From-SVN: r265600

compiler: pass a single flags argument to Backend::function

    Reviewed-on: https://go-review.googlesource.com/c/145319

* go-gcc.cc (Gcc_backend::function): Change to use a single flags
parameter.

From-SVN: r265599

compiler: add location_file

    Add support for getting the file name from a Location value.  This
    will be used by later work.

    Reviewed-on: https://go-review.googlesource.com/c/145318

* go-linemap.cc (Gcc_linemap::location_file): New method.

From-SVN: r265598

re PR tree-optimization/87785 (ICE in dr_misalignment, at tree-vectorizer.h:1245 on 454.calculix with -Ofast and -flto)

2018-10-29  Richard Biener  <rguenther@suse.de>

PR tree-optimization/87785
* tree-vect-slp.c (vect_gather_slp_loads): Only gather
internal defs.

* gcc.dg/torture/20181029-1.c: New testcase.
* gcc.dg/torture/20181029-2.c: Likewise.

From-SVN: r265596

Extract VXWORKS_NET_LIBS_RTP from VXWORKS_LIBS_RTP

2018-10-29  Olivier Hainque  <hainque@adacore.com>

* VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.

From-SVN: r265595

Introduce notion of VXWORKS_PERSONALITY

2018-10-29  Olivier Hainque  <hainque@adacore.com>

* config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
ports configuration macro, defaults to "gnu".
(VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.

From-SVN: r265594

re PR tree-optimization/87790 (ICE in vect_get_vec_def_for_operand_1, at tree-vect-stmts.c:1475)

2018-10-29  Richard Biener  <rguenther@suse.de>

PR tree-optimization/87790
* tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
(vect_make_slp_decision): Adjust.
(vect_slp_analyze_bb_1): Likewise.
(vect_detect_hybrid_slp_stmts): Properly union SLP type over
edges.

* gcc.dg/pr87790.c: New testcase.

From-SVN: r265593

[OpenACC] Support C++ "this" in OpenACC directives

2018-10-29  Joseph Myers  <joseph@codesourcery.com>
    Julian Brown  <julian@codesourcery.com>

* semantics.c (handle_omp_array_sections_1): Allow array sections with
"this" pointer for OpenACC.

Co-Authored-By: Julian Brown <julian@codesourcery.com>
From-SVN: r265591

re PR tree-optimization/87785 (ICE in dr_misalignment, at tree-vectorizer.h:1245 on 454.calculix with -Ofast and -flto)

2018-10-29  Richard Biener  <rguenther@suse.de>

PR tree-optimization/87785
* tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
and processing.
(vect_build_slp_tree): Likewise.
(vect_gather_slp_loads): New function.
(vect_analyze_slp_instance): Gather loads separately from the
SLP tree build.

From-SVN: r265588

GCOV: introduce --json-format.

2018-10-29  Martin Liska  <mliska@suse.cz>

* Makefile.in: Make dependency to json.o.
* doc/gcov.texi: Document new JSON format, remove
old intermediate format documentation.
* gcov.c (struct function_info): Come up with m_name and
m_demangled_name.
(function_info::function_info): Initialize it.
(function_info::~function_info): Release it.
(main): Rename flag_intermediate_format to flag_json_format.
(print_usage): Describe --json-format.
(process_args): Set flag_json_format.
(output_intermediate_line): Remove.
(output_intermediate_json_line): Likewise.
(get_gcov_intermediate_filename): Return new extension
".gcov.json.gz".
(output_intermediate_file): Implement JSON emission.
(output_json_intermediate_file): Implement JSON emission.
(generate_results): Use ::get_name for function name.
Handle JSON output file.
(read_graph_file): Use ::get_name instead of cplus_demangle.
(read_count_file): Likewise.
(solve_flow_graph): Likewise.
(add_line_counts): Likewise.
(accumulate_line_counts): Use new flag_json_format.
(output_function_details): Use ::get_name instead of cplus_demangle.
(output_lines): Likewise.
* json.cc (test_writing_literals): Add new tests.
* json.h (class literal): Add new boolean constructor.
2018-10-29  Martin Liska  <mliska@suse.cz>

* g++.dg/gcov/gcov-8.C: Do not check intermediate format.
* lib/gcov.exp: Remove legacy verify-intermediate.

From-SVN: r265587

Fix mingw-w64 Ada native bootstrap (PR81878).

Due to the changes in PR81878 builds of GCC8 and trunk are impossible
with Ada enabled[1][2].

The reason the patch breaks the bootstrap is due to how gnatlink receives it's
arguments.

gnatlink is usually invoked as

        $(GNATLINK) -v gnatcmd -o ../../gnat$(exeext) \
          --GCC="$(CC) $(ADA_INCLUDES)" --LINK="$(GCC_LINK)" $(TOOLS_LIBS)

so it passes $(CC) and $(GCC_LINK) as quoted arguments to the program.
Because of this quotation the msys2 shell does not translate any paths in
$(CC) and $(GCC_LINK) from their Unix version to their Windows version.

Furthermore because there are multiple paths in the values separated by space
and because the paths often contain a prefix like -L (e.g. -L/f/foo) we can't
use `fix_srcfile_path` to fix this.

An alternative solution would have been to create a stub program that echos the
commandline options it receives back, and calling this program with $(CC) and $(GCC_LINK)
unquoted to get them translated.  However this was a bit more invasive.

So instead for native compilations we add -B../../ such that it picks up the lto plugin
from the previous built compiler.  Since it's native there shouldn't be a mismatch here.

[1] https://github.com/Alexpux/MINGW-packages/pull/3877#issuecomment-408651809
[2] https://gcc.gnu.org/ml/gcc/2018-07/msg00410.html

gnattools/ChangeLog:

PR ada/81878
* Makefile.in (TOOLS_FLAGS_TO_PASS_NATIVE): Add -B ../../.

From-SVN: r265583

combine: Fix various shortcomings in make_more_copies (PR87701, PR87780)

This rewrites most of make_more_copies, in the process fixing a few PRs
and some other bugs, and working around a few target problems.  Certain
notes turn out to actually change the meaning of the RTL, so we cannot
drop them; and i386 takes subregs of hard regs.

PR rtl-optimization/87701
PR rtl-optimization/87780
* combine.c (make_more_copies): Rewrite.

From-SVN: r265582

Daily bump.

From-SVN: r265581

gimplefe-30.c: New test.

gcc/testsuite/ChangeLog:

2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>

* gcc.dg/gimplefe-30.c: New test.
* gcc.dg/gimplefe-31.c: New test.
* gcc.dg/gimplefe-32.c: New test.
* gcc.dg/gimplefe-33.c: New test.

gcc/ChangeLog:

2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>

* doc/generic.texi (ABSU_EXPR): Document.
        * match.pd (absu(x)*absu(x) -> x*x): Handle.
        (absu(absu(X)) -> absu(X)): Likewise.
        (absu(-X) -> absu(X)): Likewise.
        (absu(X)  where X is nonnegative -> X): Likewise.

From-SVN: r265578

48101_neg.cc: Remove dg-prune-output 'std' from regex pattern for versioned namespace...

2018-10-28  François Dumont  <fdumont@gcc.gnu.org>

* testsuite/23_containers/deque/48101_neg.cc: Remove dg-prune-output
'std' from regex pattern for versioned namespace compatibility.
* testsuite/23_containers/vector/48101_neg.cc: Likewise.
* testsuite/27_io/filesystem/path/io/dr2989.cc: Likewise.

From-SVN: r265575

Add D front-end, libphobos library, and D2 testsuite.

ChangeLog:

* Makefile.def (target_modules): Add libphobos.
(flags_to_pass): Add GDC, GDCFLAGS, GDC_FOR_TARGET and
GDCFLAGS_FOR_TARGET.
(dependencies): Make libphobos depend on libatomic, libbacktrace
configure, and zlib configure.
(language): Add language d.
* Makefile.in: Rebuild.
* Makefile.tpl (BUILD_EXPORTS): Add GDC and GDCFLAGS.
(HOST_EXPORTS): Add GDC.
(POSTSTAGE1_HOST_EXPORTS): Add GDC and GDC_FOR_BUILD.
(BASE_TARGET_EXPORTS): Add GDC.
(GDC_FOR_BUILD, GDC, GDCFLAGS): New variables.
(GDC_FOR_TARGET, GDC_FLAGS_FOR_TARGET): New variables.
(EXTRA_HOST_FLAGS): Add GDC.
(STAGE1_FLAGS_TO_PASS): Add GDC.
(EXTRA_TARGET_FLAGS): Add GDC and GDCFLAGS.
* config-ml.in: Treat GDC and GDCFLAGS like other compiler/flag
environment variables.
* configure: Rebuild.
* configure.ac: Add target-libphobos to target_libraries.  Set and
substitute GDC_FOR_BUILD and GDC_FOR_TARGET.

config/ChangeLog:

* multi.m4: Set GDC.

gcc/ChangeLog:

* Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
(TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
(tm_d.h, cs-tm_d.h, default-d.o): New rules.
(d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
(s-tm-texi): Also check timestamp on d-target.def.
(generated_files): Add TM_D_H and d-target-hooks-def.h.
(build/genhooks.o): Also depend on D_TARGET_DEF.
* config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
variables.
* config/aarch64/aarch64-d.c: New file.
* config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
Define.
* config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
prototype.
* config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
* config/aarch64/t-aarch64 (aarch64-d.o): New rule.
* config/arm/arm-d.c: New file.
* config/arm/arm-protos.h (arm_d_target_versions): New prototype.
* config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
* config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
* config/arm/t-arm (arm-d.o): New rule.
* config/default-d.c: New file.
* config/glibc-d.c: New file.
* config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
* config/i386/i386-d.c: New file.
* config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
* config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
* config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
(GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
* config/i386/t-i386 (i386-d.o): New rule.
* config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
* config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
* config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
* config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
* config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
* config/mips/mips-d.c: New file.
* config/mips/mips-protos.h (mips_d_target_versions): New prototype.
* config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
* config/mips/t-mips (mips-d.o): New rule.
* config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
* config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
* config/powerpcspe/powerpcspe-d.c: New file.
* config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
New prototype.
* config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
Support GNU D by using 0 as the language type.
* config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
* config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
* config/riscv/riscv-d.c: New file.
* config/riscv/riscv-protos.h (riscv_d_target_versions): New
prototype.
* config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
* config/riscv/t-riscv (riscv-d.o): New rule.
* config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
* config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
* config/rs6000/rs6000-d.c: New file.
* config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
prototype.
* config/rs6000/rs6000.c (rs6000_output_function_epilogue):
Support GNU D by using 0 as the language type.
* config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
* config/rs6000/t-rs6000 (rs6000-d.o): New rule.
* config/s390/s390-d.c: New file.
* config/s390/s390-protos.h (s390_d_target_versions): New prototype.
* config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
* config/s390/t-s390 (s390-d.o): New rule.
* config/sparc/sparc-d.c: New file.
* config/sparc/sparc-protos.h (sparc_d_target_versions): New
prototype.
* config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
* config/sparc/t-sparc (sparc-d.o): New rule.
* config/t-glibc (glibc-d.o): New rule.
* configure: Regenerated.
* configure.ac (tm_d_file): New variable.
(tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
* doc/contrib.texi (Contributors): Add self for the D frontend.
* doc/frontends.texi (G++ and GCC): Mention D as a supported language.
* doc/install.texi (Configuration): Mention libphobos as an option for
--enable-shared.  Mention d as an option for --enable-languages.
(Testing): Mention check-d as a target.
* doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
name suffixes.  Mention d as a -x option.
* doc/sourcebuild.texi (Top Level): Mention libphobos.
* doc/standards.texi (Standards): Add section on D language.
* doc/tm.texi: Regenerated.
* doc/tm.texi.in: Add @node for D language and ABI, and @hook for
TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
* dwarf2out.c (is_dlang): New function.
(gen_compile_unit_die): Use DW_LANG_D for D.
(declare_in_namespace): Return module die for D, instead of adding
extra declarations into the namespace.
(gen_namespace_die): Generate DW_TAG_module for D.
(gen_decl_die): Handle CONST_DECLSs for D.
(dwarf2out_decl): Likewise.
(prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
(prune_unused_types_walk): Handle DW_tag_interface_type same as other
kinds of aggregates.
* gcc.c (default_compilers): Add entries for .d, .dd and .di.
* genhooks.c: Include d/d-target.def.

gcc/po/ChangeLog:

* EXCLUDES: Add sources from d/dmd.

gcc/testsuite/ChangeLog:

* gcc.misc-tests/help.exp: Add D to option descriptions check.
* gdc.dg/asan/asan.exp: New file.
* gdc.dg/asan/gdc272.d: New test.
* gdc.dg/compilable.d: New test.
* gdc.dg/dg.exp: New file.
* gdc.dg/gdc254.d: New test.
* gdc.dg/gdc260.d: New test.
* gdc.dg/gdc270a.d: New test.
* gdc.dg/gdc270b.d: New test.
* gdc.dg/gdc282.d: New test.
* gdc.dg/gdc283.d: New test.
* gdc.dg/imports/gdc170.d: New test.
* gdc.dg/imports/gdc231.d: New test.
* gdc.dg/imports/gdc239.d: New test.
* gdc.dg/imports/gdc241a.d: New test.
* gdc.dg/imports/gdc241b.d: New test.
* gdc.dg/imports/gdc251a.d: New test.
* gdc.dg/imports/gdc251b.d: New test.
* gdc.dg/imports/gdc253.d: New test.
* gdc.dg/imports/gdc254a.d: New test.
* gdc.dg/imports/gdc256.d: New test.
* gdc.dg/imports/gdc27.d: New test.
* gdc.dg/imports/gdcpkg256/package.d: New test.
* gdc.dg/imports/runnable.d: New test.
* gdc.dg/link.d: New test.
* gdc.dg/lto/lto.exp: New file.
* gdc.dg/lto/ltotests_0.d: New test.
* gdc.dg/lto/ltotests_1.d: New test.
* gdc.dg/runnable.d: New test.
* gdc.dg/simd.d: New test.
* gdc.test/gdc-test.exp: New file.
* lib/gdc-dg.exp: New file.
* lib/gdc.exp: New file.

libphobos/ChangeLog:

* Makefile.am: New file.
* Makefile.in: New file.
* acinclude.m4: New file.
* aclocal.m4: New file.
* config.h.in: New file.
* configure: New file.
* configure.ac: New file.
* d_rules.am: New file.
* libdruntime/Makefile.am: New file.
* libdruntime/Makefile.in: New file.
* libdruntime/__entrypoint.di: New file.
* libdruntime/__main.di: New file.
* libdruntime/gcc/attribute.d: New file.
* libdruntime/gcc/backtrace.d: New file.
* libdruntime/gcc/builtins.d: New file.
* libdruntime/gcc/config.d.in: New file.
* libdruntime/gcc/deh.d: New file.
* libdruntime/gcc/libbacktrace.d.in: New file.
* libdruntime/gcc/unwind/arm.d: New file.
* libdruntime/gcc/unwind/arm_common.d: New file.
* libdruntime/gcc/unwind/c6x.d: New file.
* libdruntime/gcc/unwind/generic.d: New file.
* libdruntime/gcc/unwind/package.d: New file.
* libdruntime/gcc/unwind/pe.d: New file.
* m4/autoconf.m4: New file.
* m4/druntime.m4: New file.
* m4/druntime/cpu.m4: New file.
* m4/druntime/libraries.m4: New file.
* m4/druntime/os.m4: New file.
* m4/gcc_support.m4: New file.
* m4/gdc.m4: New file.
* m4/libtool.m4: New file.
* src/Makefile.am: New file.
* src/Makefile.in: New file.
* src/libgphobos.spec.in: New file.
* testsuite/Makefile.am: New file.
* testsuite/Makefile.in: New file.
* testsuite/config/default.exp: New file.
* testsuite/lib/libphobos-dg.exp: New file.
* testsuite/lib/libphobos.exp: New file.
* testsuite/testsuite_flags.in: New file.

From-SVN: r265573

re PR fortran/54613 ([F08] Add FINDLOC plus support MAXLOC/MINLOC with KIND=/BACK=)

2017-10-28  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/54613
* gfortran.h (gfc_isym_id): Add GFC_ISYM_FINDLOC.
(gfc_check_f): Add f6fl field.
(gfc_simplify_f): Add f6 field.
(gfc_resolve_f): Likewise.
(gfc_type_letter): Add optional logical_equas_int flag.
* check.c (intrinsic_type_check): New function.
(gfc_check_findloc): New function.
* intrinsics.c (gfc_type_letter): If logical_equals_int is
set, act accordingly.
(add_sym_5ml):  Reformat comment.
(add_sym_6fl): New function.
(add_functions): Add findloc.
(check_arglist): Add sixth argument, handle it.
(resolve_intrinsic): Likewise.
(check_specific): Handle findloc.
* intrinsic.h (gfc_check_findloc): Add prototype.
(gfc_simplify_findloc): Likewise.
(gfc_resolve_findloc): Likewise.
(MAX_INTRINSIC_ARGS): Adjust.
* iresolve.c (gfc_resolve_findloc): New function.
* simplify.c (gfc_simplify_minmaxloc): Make static.
(simplify_findloc_to_scalar): New function.
(simplify_findloc_nodim): New function.
(simplify_findloc_to_array): New function.
(gfc_simplify_findloc): New function.
(gfc_conv_intrinsic_findloc): New function.
(gfc_conv_intrinsic_function): Handle GFC_ISYM_FINDLOC.
(gfc_is_intrinsic_libcall): Likewise.

2017-10-28  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/54613
* Makefile.am: Add files for findloc.
* Makefile.in: Regenerated.
* libgfortran.h (gfc_array_index_type): Add.
(gfc_array_s1): Add using GFC_UINTEGER_1.
(gfc_array_s4): Likewise.
Replace unnecessary comment.
(HAVE_GFC_UINTEGER_1): Define.
(HAVE_GFC_UINTEGER_4): Define.
* m4/findloc0.m4: New file.
* m4/findloc0s.m4: New file.
* m4/findloc1.m4: New file.
* m4/findloc1s.m4: New file.
* m4/findloc2s.m4: New file.
* m4/ifindloc0.m4: New file.
* m4/ifindloc1.m4: New file.
* m4/ifindloc2.m4: New file.
* m4/iparm.m4: Use unsigned integer for characters.
        * generated/findloc0_c16.c: New file.
        * generated/findloc0_c4.c: New file.
        * generated/findloc0_c8.c: New file.
        * generated/findloc0_i1.c: New file.
        * generated/findloc0_i16.c: New file.
        * generated/findloc0_i2.c: New file.
        * generated/findloc0_i4.c: New file.
        * generated/findloc0_i8.c: New file.
        * generated/findloc0_r16.c: New file.
        * generated/findloc0_r4.c: New file.
        * generated/findloc0_r8.c: New file.
        * generated/findloc0_s1.c: New file.
        * generated/findloc0_s4.c: New file.
        * generated/findloc1_c16.c: New file.
        * generated/findloc1_c4.c: New file.
        * generated/findloc1_c8.c: New file.
        * generated/findloc1_i1.c: New file.
        * generated/findloc1_i16.c: New file.
        * generated/findloc1_i2.c: New file.
        * generated/findloc1_i4.c: New file.
        * generated/findloc1_i8.c: New file.
        * generated/findloc1_r16.c: New file.
        * generated/findloc1_r4.c: New file.
        * generated/findloc1_r8.c: New file.
        * generated/findloc1_s1.c: New file.
        * generated/findloc1_s4.c: New file.
        * generated/findloc2_s1.c: New file.
        * generated/findloc2_s4.c: New file.
        * generated/maxloc0_16_s1.c: Regenerated.
        * generated/maxloc0_16_s4.c: Regenerated.
        * generated/maxloc0_4_s1.c: Regenerated.
        * generated/maxloc0_4_s4.c: Regenerated.
        * generated/maxloc0_8_s1.c: Regenerated.
        * generated/maxloc0_8_s4.c: Regenerated.
        * generated/maxloc1_16_s1.c: Regenerated.
        * generated/maxloc1_16_s4.c: Regenerated.
        * generated/maxloc1_4_s1.c: Regenerated.
        * generated/maxloc1_4_s4.c: Regenerated.
        * generated/maxloc1_8_s1.c: Regenerated.
        * generated/maxloc1_8_s4.c: Regenerated.
        * generated/maxloc2_16_s1.c: Regenerated.
        * generated/maxloc2_16_s4.c: Regenerated.
        * generated/maxloc2_4_s1.c: Regenerated.
        * generated/maxloc2_4_s4.c: Regenerated.
        * generated/maxloc2_8_s1.c: Regenerated.
        * generated/maxloc2_8_s4.c: Regenerated.
        * generated/maxval0_s1.c: Regenerated.
        * generated/maxval0_s4.c: Regenerated.
        * generated/maxval1_s1.c: Regenerated.
        * generated/maxval1_s4.c: Regenerated.
        * generated/minloc0_16_s1.c: Regenerated.
        * generated/minloc0_16_s4.c: Regenerated.
        * generated/minloc0_4_s1.c: Regenerated.
        * generated/minloc0_4_s4.c: Regenerated.
        * generated/minloc0_8_s1.c: Regenerated.
        * generated/minloc0_8_s4.c: Regenerated.
        * generated/minloc1_16_s1.c: Regenerated.
        * generated/minloc1_16_s4.c: Regenerated.
        * generated/minloc1_4_s1.c: Regenerated.
        * generated/minloc1_4_s4.c: Regenerated.
        * generated/minloc1_8_s1.c: Regenerated.
        * generated/minloc1_8_s4.c: Regenerated.
        * generated/minloc2_16_s1.c: Regenerated.
        * generated/minloc2_16_s4.c: Regenerated.
        * generated/minloc2_4_s1.c: Regenerated.
        * generated/minloc2_4_s4.c: Regenerated.
        * generated/minloc2_8_s1.c: Regenerated.
        * generated/minloc2_8_s4.c: Regenerated.
        * generated/minval0_s1.c: Regenerated.
        * generated/minval0_s4.c: Regenerated.
        * generated/minval1_s1.c: Regenerated.
        * generated/minval1_s4.c: Regenerated.

2017-10-28  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/54613
* gfortran.dg/findloc_1.f90: New test.
* gfortran.dg/findloc_2.f90: New test.
* gfortran.dg/findloc_3.f90: New test.
* gfortran.dg/findloc_4.f90: New test.
* gfortran.dg/findloc_5.f90: New test.
* gfortran.dg/findloc_6.f90: New test.

From-SVN: r265570

darwin - fix powerpc-darwin stack alignments

2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>

PR target/85669
* config/rs6000/darwin.h (STACK_BOUNDARY): New.
(RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
(STACK_DYNAMIC_OFFSET): Likewise.

From-SVN: r265568

Daily bump.

From-SVN: r265564

re PR target/80024 (nios2: unclear wording "numeric digits" in diagnostic)

2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>

PR target/80024

gcc/
* config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
error message.

From-SVN: r265561

re PR fortran/86907 (bogus warning "No location in expression near")

2018-10-27  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/86907
* frontend-passes.c (check_locus_code): Add information that
warning about missing location information points to an
inconsisten internal state.
(check_locus_expr): Likewise.

From-SVN: r265559

Daily bump.

From-SVN: r265554

powerpc: Fix typos in the manual

2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>

* doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
return type and other typos.

From-SVN: r265548

rs6000-string.c (expand_strncmp_gpr_sequence): Change to a shorter sequence with fewer branches.

2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>

* config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
a shorter sequence with fewer branches.
(emit_final_str_compare_gpr): Ditto.

From-SVN: r265546

sse-addss-1.c: Call abort under DEBUG also; formatting cleanup.

2018-10-26  Bill Schmidt  <wschmidt@linux.ibm.com>
    Jinsong Ji  <jji@us.ibm.com>

* gcc.target/powerpc/sse-addss-1.c: Call abort under DEBUG also;
formatting cleanup.
* gcc.target/powerpc/sse-pavgw-1.c: Likewise.
* gcc.target/powerpc/sse2-addsd-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtdq2pd-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtpd2dq-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtpd2ps-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtps2dq-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtps2pd-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtsd2si-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtsd2si-2.c: Likewise.
* gcc.target/powerpc/sse2-cvtsd2ss-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtss2sd-1.c: Likewise.
* gcc.target/powerpc/sse2-cvttpd2dq-1.c: Likewise.
* gcc.target/powerpc/sse2-cvttsd2si-1.c: Likewise.
* gcc.target/powerpc/sse2-divpd-1.c: Likewise.
* gcc.target/powerpc/sse2-divsd-1.c: Likewise.
* gcc.target/powerpc/sse2-maxsd-1.c: Likewise.
* gcc.target/powerpc/sse2-minsd-1.c: Likewise.
* gcc.target/powerpc/sse2-movmskpd-1.c: Likewise.
* gcc.target/powerpc/sse2-movq-1.c: Likewise.
* gcc.target/powerpc/sse2-movsd-3.c: Likewise.
* gcc.target/powerpc/sse2-mulpd-1.c: Likewise.
* gcc.target/powerpc/sse2-mulsd-1.c: Likewise.
* gcc.target/powerpc/sse2-packssdw-1.c: Likewise.
* gcc.target/powerpc/sse2-packsswb-1.c: Likewise.
* gcc.target/powerpc/sse2-packuswb-1.c: Likewise.
* gcc.target/powerpc/sse2-paddsb-1.c: Likewise.
* gcc.target/powerpc/sse2-paddsw-1.c: Likewise.
* gcc.target/powerpc/sse2-paddusb-1.c: Likewise.
* gcc.target/powerpc/sse2-pmovmskb-1.c: Likewise.
* gcc.target/powerpc/sse2-pmulhw-1.c: Likewise.
* gcc.target/powerpc/sse2-pmuludq1.c: Likewise.
* gcc.target/powerpc/sse2-psadbw-1.c: Likewise.
* gcc.target/powerpc/sse2-pshufd-1.c: Likewise.
* gcc.target/powerpc/sse2-pshufhw-1.c: Likewise.
* gcc.target/powerpc/sse2-pshuflw-1.c: Likewise.
* gcc.target/powerpc/sse2-pslld-2.c: Likewise.
* gcc.target/powerpc/sse2-pslldq-1.c: Likewise.
* gcc.target/powerpc/sse2-psrld-1.c: Likewise.
* gcc.target/powerpc/sse2-psrld-2.c: Likewise.
* gcc.target/powerpc/sse2-psrldq-1.c: Likewise.
* gcc.target/powerpc/sse2-psubusb-1.c: Likewise.
* gcc.target/powerpc/sse2-sqrtpd-1.c: Likewise.
* gcc.target/powerpc/sse2-subsd-1.c: Likewise.

Co-Authored-By: Jinsong Ji <jji@us.ibm.com>
From-SVN: r265545

[rs6000] Add tests for compatible implementations of x86 SSSE3 intrinsics

This is part 2/2 for contributing PPC64LE support for X86 SSSE3 instrisics.
This patch includes testsuite/gcc.target tests, copied from gcc.target/i386,
for the intrinsics defined in tmmintrin.h.

[gcc/testsuite]

2018-10-26  Paul A. Clarke  <pc@us.ibm.com>

* gcc.target/powerpc/ssse3-check.h: New file.
* gcc.target/powerpc/ssse3-vals.h: New file.
* gcc.target/powerpc/ssse3-pabsb.c: New file.
* gcc.target/powerpc/ssse3-pabsd.c: New file.
* gcc.target/powerpc/ssse3-pabsw.c: New file.
* gcc.target/powerpc/ssse3-palignr.c: New file.
* gcc.target/powerpc/ssse3-phaddd.c: New file.
* gcc.target/powerpc/ssse3-phaddsw.c: New file.
* gcc.target/powerpc/ssse3-phaddw.c: New file.
* gcc.target/powerpc/ssse3-phsubd.c: New file.
* gcc.target/powerpc/ssse3-phsubsw.c: New file.
* gcc.target/powerpc/ssse3-phsubw.c: New file.
* gcc.target/powerpc/ssse3-pmaddubsw.c: New file.
* gcc.target/powerpc/ssse3-pmulhrsw.c: New file.
* gcc.target/powerpc/ssse3-pshufb.c: New file.
* gcc.target/powerpc/ssse3-psignb.c: New file.
* gcc.target/powerpc/ssse3-psignd.c: New file.
* gcc.target/powerpc/ssse3-psignw.c: New file.

From-SVN: r265544

[rs6000] Add compatible implementations of x86 SSSE3 intrinsics

This is a follow-on to earlier commits for adding compatibility
implementations of x86 intrinsics for PPC64LE.  This is the first of
two patches.  This patch adds the 32 x86 intrinsics from
<tmmintrin.h> ("SSSE3").  (Patch 2/2 adds tests for these intrinsics,
and briefly describes the tests performed.)

gcc/ChangeLog:

2018-10-26  Paul A. Clarke  <pc@us.ibm.com>

* config/rs6000/tmmintrin.h: New file.
* config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.

From-SVN: r265542

libgo: simplify gotest script to avoid sed substitution to \n

    Reviewed-on: https://go-review.googlesource.com/c/145057

From-SVN: r265541

[rs6000] Enable 32bit support for tests of x86-compatibile intrinsics

Also, bugfix for mmx-packuswb-1.c, which had ommitted the required
"-mpower8-vector" from dg-options.

2018-10-26  Paul A. Clarke  <pc@us.ibm.com>

gcc/testsuite/ChangeLog:

* gcc.target/powerpc/mmx-packs.c: Enable 32 bit execution.
* gcc.target/powerpc/mmx-packssdw-1.c: Likewise.
* gcc.target/powerpc/mmx-packsswb-1.c: Likewise.
* gcc.target/powerpc/mmx-packs.c: Likewise.
* gcc.target/powerpc/mmx-packssdw-1.c: Likewise.
* gcc.target/powerpc/mmx-packsswb-1.c: Likewise.
* gcc.target/powerpc/mmx-paddb-1.c: Likewise.
* gcc.target/powerpc/mmx-paddd-1.c: Likewise.
* gcc.target/powerpc/mmx-paddsb-1.c: Likewise.
* gcc.target/powerpc/mmx-paddsw-1.c: Likewise.
* gcc.target/powerpc/mmx-paddusb-1.c: Likewise.
* gcc.target/powerpc/mmx-paddusw-1.c: Likewise.
* gcc.target/powerpc/mmx-paddw-1.c: Likewise.
* gcc.target/powerpc/mmx-pcmpeqb-1.c: Likewise.
* gcc.target/powerpc/mmx-pcmpeqd-1.c: Likewise.
* gcc.target/powerpc/mmx-pcmpeqw-1.c: Likewise.
* gcc.target/powerpc/mmx-pcmpgtb-1.c: Likewise.
* gcc.target/powerpc/mmx-pcmpgtd-1.c: Likewise.
* gcc.target/powerpc/mmx-pcmpgtw-1.c: Likewise.
* gcc.target/powerpc/mmx-pmaddwd-1.c: Likewise.
* gcc.target/powerpc/mmx-pmulhw-1.c: Likewise.
* gcc.target/powerpc/mmx-pmullw-1.c: Likewise.
* gcc.target/powerpc/mmx-pslld-1.c: Likewise.
* gcc.target/powerpc/mmx-psllw-1.c: Likewise.
* gcc.target/powerpc/mmx-psrad-1.c: Likewise.
* gcc.target/powerpc/mmx-psraw-1.c: Likewise.
* gcc.target/powerpc/mmx-psrld-1.c: Likewise.
* gcc.target/powerpc/mmx-psrlw-1.c: Likewise.
* gcc.target/powerpc/mmx-psubb-2.c: Likewise.
* gcc.target/powerpc/mmx-psubd-2.c: Likewise.
* gcc.target/powerpc/mmx-psubsb-1.c: Likewise.
* gcc.target/powerpc/mmx-psubsw-1.c: Likewise.
* gcc.target/powerpc/mmx-psubusb-1.c: Likewise.
* gcc.target/powerpc/mmx-psubusw-1.c: Likewise.
* gcc.target/powerpc/mmx-psubw-2.c: Likewise.
* gcc.target/powerpc/mmx-punpckhbw-1.c: Likewise.
* gcc.target/powerpc/mmx-punpckhdq-1.c: Likewise.
* gcc.target/powerpc/mmx-punpckhwd-1.c: Likewise.
* gcc.target/powerpc/mmx-punpcklbw-1.c: Likewise.
* gcc.target/powerpc/mmx-punpckldq-1.c: Likewise.
* gcc.target/powerpc/mmx-punpcklwd-1.c: Likewise.
* gcc.target/powerpc/pr37191.c: Likewise.
* gcc.target/powerpc/sse-addps-1.c: Likewise.
* gcc.target/powerpc/sse-addss-1.c: Likewise.
* gcc.target/powerpc/sse-andnps-1.c: Likewise.
* gcc.target/powerpc/sse-andps-1.c: Likewise.
* gcc.target/powerpc/sse-cmpss-1.c: Likewise.
* gcc.target/powerpc/sse-cvtpi16ps-1.c: Likewise.
* gcc.target/powerpc/sse-cvtpi32ps-1.c: Likewise.
* gcc.target/powerpc/sse-cvtpi32x2ps-1.c: Likewise.
* gcc.target/powerpc/sse-cvtpi8ps-1.c: Likewise.
* gcc.target/powerpc/sse-cvtpspi16-1.c: Likewise.
* gcc.target/powerpc/sse-cvtpspi8-1.c: Likewise.
* gcc.target/powerpc/sse-cvtpu16ps-1.c: Likewise.
* gcc.target/powerpc/sse-cvtpu8ps-1.c: Likewise.
* gcc.target/powerpc/sse-cvtsi2ss-1.c: Likewise.
* gcc.target/powerpc/sse-cvtsi2ss-2.c: Likewise.
* gcc.target/powerpc/sse-cvtss2si-1.c: Likewise.
* gcc.target/powerpc/sse-cvtss2si-2.c: Likewise.
* gcc.target/powerpc/sse-cvttss2si-1.c: Likewise.
* gcc.target/powerpc/sse-cvttss2si-2.c: Likewise.
* gcc.target/powerpc/sse-divps-1.c: Likewise.
* gcc.target/powerpc/sse-divss-1.c: Likewise.
* gcc.target/powerpc/sse-maxps-1.c: Likewise.
* gcc.target/powerpc/sse-maxps-2.c: Likewise.
* gcc.target/powerpc/sse-maxss-1.c: Likewise.
* gcc.target/powerpc/sse-minps-1.c: Likewise.
* gcc.target/powerpc/sse-minps-2.c: Likewise.
* gcc.target/powerpc/sse-minss-1.c: Likewise.
* gcc.target/powerpc/sse-movaps-1.c: Likewise.
* gcc.target/powerpc/sse-movaps-2.c: Likewise.
* gcc.target/powerpc/sse-movhlps-1.c: Likewise.
* gcc.target/powerpc/sse-movhps-1.c: Likewise.
* gcc.target/powerpc/sse-movhps-2.c: Likewise.
* gcc.target/powerpc/sse-movlhps-1.c: Likewise.
* gcc.target/powerpc/sse-movlps-1.c: Likewise.
* gcc.target/powerpc/sse-movlps-2.c: Likewise.
* gcc.target/powerpc/sse-movmskb-1.c: Likewise.
* gcc.target/powerpc/sse-movmskps-1.c: Likewise.
* gcc.target/powerpc/sse-movss-1.c: Likewise.
* gcc.target/powerpc/sse-movss-2.c: Likewise.
* gcc.target/powerpc/sse-movss-3.c: Likewise.
* gcc.target/powerpc/sse-mulps-1.c: Likewise.
* gcc.target/powerpc/sse-mulss-1.c: Likewise.
* gcc.target/powerpc/sse-orps-1.c: Likewise.
* gcc.target/powerpc/sse-pavgw-1.c: Likewise.
* gcc.target/powerpc/sse-pmaxsw-1.c: Likewise.
* gcc.target/powerpc/sse-pmaxub-1.c: Likewise.
* gcc.target/powerpc/sse-pminsw-1.c: Likewise.
* gcc.target/powerpc/sse-pminub-1.c: Likewise.
* gcc.target/powerpc/sse-pmulhuw-1.c: Likewise.
* gcc.target/powerpc/sse-psadbw-1.c: Likewise.
* gcc.target/powerpc/sse-rcpps-1.c: Likewise.
* gcc.target/powerpc/sse-rsqrtps-1.c: Likewise.
* gcc.target/powerpc/sse-shufps-1.c: Likewise.
* gcc.target/powerpc/sse-sqrtps-1.c: Likewise.
* gcc.target/powerpc/sse-subps-1.c: Likewise.
* gcc.target/powerpc/sse-subss-1.c: Likewise.
* gcc.target/powerpc/sse-ucomiss-1.c: Likewise.
* gcc.target/powerpc/sse-ucomiss-2.c: Likewise.
* gcc.target/powerpc/sse-ucomiss-3.c: Likewise.
* gcc.target/powerpc/sse-ucomiss-4.c: Likewise.
* gcc.target/powerpc/sse-ucomiss-5.c: Likewise.
* gcc.target/powerpc/sse-ucomiss-6.c: Likewise.
* gcc.target/powerpc/sse-unpckhps-1.c: Likewise.
* gcc.target/powerpc/sse-unpcklps-1.c: Likewise.
* gcc.target/powerpc/sse-xorps-1.c: Likewise.
* gcc.target/powerpc/sse2-addpd-1.c: Likewise.
* gcc.target/powerpc/sse2-addsd-1.c: Likewise.
* gcc.target/powerpc/sse2-andnpd-1.c: Likewise.
* gcc.target/powerpc/sse2-andpd-1.c: Likewise.
* gcc.target/powerpc/sse2-cmppd-1.c: Likewise.
* gcc.target/powerpc/sse2-cmpsd-1.c: Likewise.
* gcc.target/powerpc/sse2-comisd-1.c: Likewise.
* gcc.target/powerpc/sse2-comisd-2.c: Likewise.
* gcc.target/powerpc/sse2-comisd-3.c: Likewise.
* gcc.target/powerpc/sse2-comisd-4.c: Likewise.
* gcc.target/powerpc/sse2-comisd-5.c: Likewise.
* gcc.target/powerpc/sse2-comisd-6.c: Likewise.
* gcc.target/powerpc/sse2-cvtdq2pd-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtdq2ps-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtpd2dq-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtpd2ps-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtps2dq-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtps2pd-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtsd2si-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtsd2si-2.c: Likewise.
* gcc.target/powerpc/sse2-cvtsd2ss-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtsi2sd-1.c: Likewise.
* gcc.target/powerpc/sse2-cvtsi2sd-2.c: Likewise.
* gcc.target/powerpc/sse2-cvtss2sd-1.c: Likewise.
* gcc.target/powerpc/sse2-cvttpd2dq-1.c: Likewise.
* gcc.target/powerpc/sse2-cvttps2dq-1.c: Likewise.
* gcc.target/powerpc/sse2-cvttsd2si-1.c: Likewise.
* gcc.target/powerpc/sse2-cvttsd2si-2.c: Likewise.
* gcc.target/powerpc/sse2-divpd-1.c: Likewise.
* gcc.target/powerpc/sse2-divsd-1.c: Likewise.
* gcc.target/powerpc/sse2-maxpd-1.c: Likewise.
* gcc.target/powerpc/sse2-maxsd-1.c: Likewise.
* gcc.target/powerpc/sse2-minpd-1.c: Likewise.
* gcc.target/powerpc/sse2-minsd-1.c: Likewise.
* gcc.target/powerpc/sse2-mmx.c: Likewise.
* gcc.target/powerpc/sse2-movhpd-1.c: Likewise.
* gcc.target/powerpc/sse2-movhpd-2.c: Likewise.
* gcc.target/powerpc/sse2-movlpd-1.c: Likewise.
* gcc.target/powerpc/sse2-movlpd-2.c: Likewise.
* gcc.target/powerpc/sse2-movmskpd-1.c: Likewise.
* gcc.target/powerpc/sse2-movq-1.c: Likewise.
* gcc.target/powerpc/sse2-movq-2.c: Likewise.
* gcc.target/powerpc/sse2-movq-3.c: Likewise.
* gcc.target/powerpc/sse2-movsd-1.c: Likewise.
* gcc.target/powerpc/sse2-movsd-2.c: Likewise.
* gcc.target/powerpc/sse2-movsd-3.c: Likewise.
* gcc.target/powerpc/sse2-mulpd-1.c: Likewise.
* gcc.target/powerpc/sse2-mulsd-1.c: Likewise.
* gcc.target/powerpc/sse2-orpd-1.c: Likewise.
* gcc.target/powerpc/sse2-packssdw-1.c: Likewise.
* gcc.target/powerpc/sse2-packsswb-1.c: Likewise.
* gcc.target/powerpc/sse2-packuswb-1.c: Likewise.
* gcc.target/powerpc/sse2-paddb-1.c: Likewise.
* gcc.target/powerpc/sse2-paddd-1.c: Likewise.
* gcc.target/powerpc/sse2-paddq-1.c: Likewise.
* gcc.target/powerpc/sse2-paddsb-1.c: Likewise.
* gcc.target/powerpc/sse2-paddsw-1.c: Likewise.
* gcc.target/powerpc/sse2-paddusb-1.c: Likewise.
* gcc.target/powerpc/sse2-paddusw-1.c: Likewise.
* gcc.target/powerpc/sse2-paddw-1.c: Likewise.
* gcc.target/powerpc/sse2-pand-1.c: Likewise.
* gcc.target/powerpc/sse2-pandn-1.c: Likewise.
* gcc.target/powerpc/sse2-pavgb-1.c: Likewise.
* gcc.target/powerpc/sse2-pavgw-1.c: Likewise.
* gcc.target/powerpc/sse2-pcmpeqb-1.c: Likewise.
* gcc.target/powerpc/sse2-pcmpeqd-1.c: Likewise.
* gcc.target/powerpc/sse2-pcmpeqw-1.c: Likewise.
* gcc.target/powerpc/sse2-pcmpgtb-1.c: Likewise.
* gcc.target/powerpc/sse2-pcmpgtd-1.c: Likewise.
* gcc.target/powerpc/sse2-pcmpgtw-1.c: Likewise.
* gcc.target/powerpc/sse2-pextrw.c: Likewise.
* gcc.target/powerpc/sse2-pinsrw.c: Likewise.
* gcc.target/powerpc/sse2-pmaddwd-1.c: Likewise.
* gcc.target/powerpc/sse2-pmaxsw-1.c: Likewise.
* gcc.target/powerpc/sse2-pmaxub-1.c: Likewise.
* gcc.target/powerpc/sse2-pminsw-1.c: Likewise.
* gcc.target/powerpc/sse2-pminub-1.c: Likewise.
* gcc.target/powerpc/sse2-pmovmskb-1.c: Likewise.
* gcc.target/powerpc/sse2-pmulhuw-1.c: Likewise.
* gcc.target/powerpc/sse2-pmulhw-1.c: Likewise.
* gcc.target/powerpc/sse2-pmullw-1.c: Likewise.
* gcc.target/powerpc/sse2-pmuludq-1.c: Likewise.
* gcc.target/powerpc/sse2-por-1.c: Likewise.
* gcc.target/powerpc/sse2-psadbw-1.c: Likewise.
* gcc.target/powerpc/sse2-pshufd-1.c: Likewise.
* gcc.target/powerpc/sse2-pshufhw-1.c: Likewise.
* gcc.target/powerpc/sse2-pshuflw-1.c: Likewise.
* gcc.target/powerpc/sse2-pslld-1.c: Likewise.
* gcc.target/powerpc/sse2-pslld-2.c: Likewise.
* gcc.target/powerpc/sse2-pslldq-1.c: Likewise.
* gcc.target/powerpc/sse2-psllq-1.c: Likewise.
* gcc.target/powerpc/sse2-psllq-2.c: Likewise.
* gcc.target/powerpc/sse2-psllw-1.c: Likewise.
* gcc.target/powerpc/sse2-psllw-2.c: Likewise.
* gcc.target/powerpc/sse2-psrad-1.c: Likewise.
* gcc.target/powerpc/sse2-psrad-2.c: Likewise.
* gcc.target/powerpc/sse2-psraw-1.c: Likewise.
* gcc.target/powerpc/sse2-psraw-2.c: Likewise.
* gcc.target/powerpc/sse2-psrld-1.c: Likewise.
* gcc.target/powerpc/sse2-psrld-2.c: Likewise.
* gcc.target/powerpc/sse2-psrldq-1.c: Likewise.
* gcc.target/powerpc/sse2-psrlq-1.c: Likewise.
* gcc.target/powerpc/sse2-psrlq-2.c: Likewise.
* gcc.target/powerpc/sse2-psrlw-1.c: Likewise.
* gcc.target/powerpc/sse2-psrlw-2.c: Likewise.
* gcc.target/powerpc/sse2-psubb-1.c: Likewise.
* gcc.target/powerpc/sse2-psubd-1.c: Likewise.
* gcc.target/powerpc/sse2-psubq-1.c: Likewise.
* gcc.target/powerpc/sse2-psubsb-1.c: Likewise.
* gcc.target/powerpc/sse2-psubsw-1.c: Likewise.
* gcc.target/powerpc/sse2-psubusb-1.c: Likewise.
* gcc.target/powerpc/sse2-psubusw-1.c: Likewise.
* gcc.target/powerpc/sse2-psubw-1.c: Likewise.
* gcc.target/powerpc/sse2-punpckhbw-1.c: Likewise.
* gcc.target/powerpc/sse2-punpckhdq-1.c: Likewise.
* gcc.target/powerpc/sse2-punpckhqdq-1.c: Likewise.
* gcc.target/powerpc/sse2-punpckhwd-1.c: Likewise.
* gcc.target/powerpc/sse2-punpcklbw-1.c: Likewise.
* gcc.target/powerpc/sse2-punpckldq-1.c: Likewise.
* gcc.target/powerpc/sse2-punpcklqdq-1.c: Likewise.
* gcc.target/powerpc/sse2-punpcklwd-1.c: Likewise.
* gcc.target/powerpc/sse2-pxor-1.c: Likewise.
* gcc.target/powerpc/sse2-shufpd-1.c: Likewise.
* gcc.target/powerpc/sse2-sqrtpd-1.c: Likewise.
* gcc.target/powerpc/sse2-subpd-1.c: Likewise.
* gcc.target/powerpc/sse2-subsd-1.c: Likewise.
* gcc.target/powerpc/sse2-ucomisd-1.c: Likewise.
* gcc.target/powerpc/sse2-ucomisd-2.c: Likewise.
* gcc.target/powerpc/sse2-ucomisd-3.c: Likewise.
* gcc.target/powerpc/sse2-ucomisd-4.c: Likewise.
* gcc.target/powerpc/sse2-ucomisd-5.c: Likewise.
* gcc.target/powerpc/sse2-ucomisd-6.c: Likewise.
* gcc.target/powerpc/sse2-unpckhpd-1.c: Likewise.
* gcc.target/powerpc/sse2-unpcklpd-1.c: Likewise.
* gcc.target/powerpc/sse2-xorpd-1.c: Likewise.
* gcc.target/powerpc/sse3-addsubpd.c: Likewise.
* gcc.target/powerpc/sse3-addsubps.c: Likewise.
* gcc.target/powerpc/sse3-haddpd.c: Likewise.
* gcc.target/powerpc/sse3-haddps.c: Likewise.
* gcc.target/powerpc/sse3-hsubpd.c: Likewise.
* gcc.target/powerpc/sse3-hsubps.c: Likewise.
* gcc.target/powerpc/sse3-lddqu.c: Likewise.
* gcc.target/powerpc/sse3-movddup.c: Likewise.
* gcc.target/powerpc/sse3-movshdup.c: Likewise.
* gcc.target/powerpc/sse3-movsldup.c: Likewise.
* gcc.target/powerpc/mmx-packuswb-1.c: Likewise. Also, add
ommitted "-mpower8-vector" to dg-options.

From-SVN: r265537

[rs6000] x86 vector intrinsics compatibility: clean-ups for 32bit support

Implement various corrections in the compatibility implementations of the
x86 vector intrinsics found after enabling 32bit mode for the associated
test cases.  (Actual enablement coming in a subsequent patch.)

2018-10-26  Paul A. Clarke  <pc@us.ibm.com>

gcc/ChangeLog:

* config/rs6000/mmintrin.h: Enable 32bit compilation.
* config/rs6000/xmmintrin.h: Likewise.

From-SVN: r265535

libgo: avoid use of 'local' directive in shell script

    Avoid declaring shell variables with 'local' (not supported
    on all systems of interest).

    Reviewed-on: https://go-review.googlesource.com/c/145021

From-SVN: r265534

libgo: fix improperly mangled linker symbol directive

    Fix asm name directive for the C version of log/syslog.syslog_c,
    which didn't get included in the recent name mangling change.

    Reviewed-on: https://go-review.googlesource.com/c/145017

From-SVN: r265533

[rs6000] Fix _mm_extract_pi16 for big-endian

For compatibility implementation of x86 vector intrinsic, _mm_extract_pi16,
adjust shift value for big-endian mode.

gcc/ChangeLog:

2018-10-25  Paul A. Clarke  <pc@us.ibm.com>

* config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.

From-SVN: r265531

tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set and wrapper.

2018-10-26  Richard Biener  <rguenther@suse.de>

* tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
and wrapper.
(vect_mark_slp_stmts_relevant): Likewise.
(vect_detect_hybrid_slp_stmts): Likewise.
(vect_bb_slp_scalar_cost): Likewise.
(vect_remove_slp_scalar_calls): Likewise.

From-SVN: r265528

update_version_svn (IGNORE_BRANCHES): Add gcc-6-branch.

* update_version_svn (IGNORE_BRANCHES): Add gcc-6-branch.
* crontab: Remove gcc-6-branch entry.

From-SVN: r265527

gcc_release (error, inform): Use $@ instead of $1.

* gcc_release (error, inform): Use $@ instead of $1.
(build_sources): Check for ^[[:blank:]]*GCC in both index.html
and changes.html, rather than for GCC in one and ^GCC in another one.

From-SVN: r265526

ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.

* ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
(warn_types_mismatch): Fix walk of DECL_NAME.
(odr_types_equivalent_p): Fix overactive assert.
* lto/lto-symtab.c (lto_symtab_merge_decls_2): Fix extra space.

* g++.dg/lto/odr-1_0.C: Fix template.
* g++.dg/lto/odr-1_1.C: Fix template.

From-SVN: r265523

re PR tree-optimization/87105 (Autovectorization [X86, SSE2, AVX2, DoublePrecision])

2018-10-26  Richard Biener  <rguenther@suse.de>

PR tree-optimization/87105
* tree-vectorizer.h (_slp_tree::refcnt): New member.
* tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
refcnt.
(vect_create_new_slp_node): Initialize refcnt to one.
(bst_traits): Move.
(scalar_stmts_set_t, bst_fail): Remove.
(vect_build_slp_tree_2): Add bst_map argument and adjust calls.
(vect_build_slp_tree): Add bst_map argument and lookup
already created SLP nodes.
(vect_print_slp_tree): Handle a SLP graph, print SLP node
addresses.
(vect_slp_rearrange_stmts): Handle a SLP graph.
(vect_analyze_slp_instance): Adjust and free SLP nodes from
the CSE map.  Fix indenting.
(vect_schedule_slp_instance): Add short-cut.

* g++.dg/vect/slp-pr87105.cc: Adjust.
* gcc.dg/torture/20181024-1.c: New testcase.
* g++.dg/opt/20181025-1.C: Likewise.

From-SVN: r265522

Relax hash function to match equals function behavior (PR testsuite/86158).

2018-10-26  Martin Liska  <mliska@suse.cz>

PR testsuite/86158
* ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
addr_expr and not with pointers.

From-SVN: r265521

* tree.c (free_lang_data_in_type): Only check main variants.

From-SVN: r265520

ipa-devirt.c (warn_odr): Make static.

* ipa-devirt.c (warn_odr): Make static.
(types_same_for_odr): Drop strict variant.
(types_odr_comparable): Likewise.
(odr_or_derived_type_p): Look for main variants.
(odr_name_hasher::equal): Cleanup comment.
(odr_subtypes_equivalent): Add warn and warned arguments; check main
variants.
(type_variants_equivalent_p): break out from ...
(odr_types_equivalent): ... here; go for main variants where needed.
(warn_odr): ... here; turn static.
(warn_types_mismatch): Compare mangled names of main variants.
* ipa-utils.h (types_odr_comparable): Drop strict parameter.
(type_with_linkage_p): Sanity check that we look at main variant.
* lto.c (lto_read_decls): Only consider main variant to be ODR type.
* tree.h (types_same_for_odr): Drop strict argument.

From-SVN: r265519

re PR lto/87754 (ICE in odr_types_equivalent_p, at ipa-devirt.c:1250)

2018-10-26  Richard Biener  <rguenther@suse.de>

PR testsuite/87754
* g++.dg/lto/odr-1_0.C: Fix pattern.

From-SVN: r265518

re PR tree-optimization/87746 (ICE in vect_update_misalignment_for_peel, at tree-vect-data-refs.c:1035)

2018-10-26  Richard Biener  <rguenther@suse.de>

PR tree-optimization/87746
* tree-vect-data-refs.c (vect_update_misalignment_for_peel):
Simplify and fix WRT strided store groups with size not
equal to step in element count.
(vect_analyze_group_access_1): Dump the whole group.

* gcc.dg/pr87746.c: New testcase.

From-SVN: r265517

Don't xfail gcc.dg/pr78973-2.c on non-ilp64 targets

* gcc.dg/pr78973-2.c: Remove xfail on dg-warning.

From-SVN: r265516

libgo: don't use wc in gotest

    The wc command is not in the GNU approved list of Makefile utilities
    (https://www.gnu.org/prep/standards/html_node/Utilities-in-Makefiles.html#Utilities-in-Makefiles).

    Reviewed-on: https://go-review.googlesource.com/c/144897

From-SVN: r265515

Daily bump.

From-SVN: r265514

compiler: improve name mangling for packpaths

    The current implementation of Gogo::pkgpath_for_symbol was written in
    a way that allowed two distinct package paths to map to the same
    symbol, which could cause collisions at link- time or compile-time.

    Switch to a better mangling scheme to insure that we get a unique
    packagepath symbol for each package. In the new scheme instead of having
    separate mangling schemes for identifiers and package paths, the
    main identifier mangler ("go_encode_id") now handles mangling of
    both packagepath characters and identifier characters.

    The new mangling scheme is more intrusive: "foo/bar.Baz" is mangled as
    "foo..z2fbar.Baz" instead of "foo_bar.Baz". To mitigate this, this
    patch also adds a demangling capability so that function names
    returned from runtime.CallersFrames are converted back to their
    original unmangled form.

    Changing the pkgpath_for_symbol scheme requires updating a number of
    //go:linkname directives and C "__asm__" directives to match the new
    scheme, as well as updating the 'gotest' driver (which makes
    assumptions about the correct mapping from pkgpath symbol to package
    name).

    Fixes golang/go#27534.

    Reviewed-on: https://go-review.googlesource.com/c/135455

From-SVN: r265510

rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT, [...]): Rename base overloaded name.

gcc/ChangeLog:

2018-10-25  Carl Love  <cel@us.ibm.com>

* config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
precicion entry for each overloaded builtin.
* config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
VSCEDPUO): Rename overloaded name.
(VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
* config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
define_expand for xscmpexqp instruction.
(*xscmpexpqp): Add define_insn for the xscmpexqp instruction.

gcc/testsuite/ChangeLog:

2018-10-25  Carl Love  <cel@us.ibm.com>

* gcc.target/powerpc/float128-cmp2-runnable.c: New test file.

From-SVN: r265509

emmintrin.h (_mm_slli_epi16): Replace deprecated function with vec_sl.

2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
    Jinsong Ji <jji@us.ibm.com>

* config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
function with vec_sl.
(_mm_slli_epi32): Likewise.
(_mm_slli_epi64): Likewise.
(_mm_srai_epi16): Replace deprecated function with vec_sra.
(_mm_srai_epi32): Likewise.
(_mm_srli_epi16): Replace deprecated function with vec_sr.
(_mm_srli_epi32): Likewise.
(_mm_srli_epi64): Likewise.
(_mm_sll_epi16): Replace deprecated function with vec_sl.
(_mm_sll_epi32): Likewise.
(_mm_sll_epi64): Likewise.
(_mm_sra_epi16): Replace deprecated function with vec_sra.
(_mm_sra_epi32): Likewise.
(_mm_srl_epi16): Replace deprecated function with vec_sr.
(_mm_srl_epi32): Likewise.
(_mm_srl_epi64): Likewise.

Co-Authored-By: Jinsong Ji <jji@us.ibm.com>
From-SVN: r265508

emmintrin.h (_mm_sll_epi16): Replace comparison operators with vec_cmp* for compatibility due to unfortunate...

2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
    Jinsong Ji <jji@us.ibm.com>

* gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
comparison operators with vec_cmp* for compatibility due to
unfortunate history; clean up formatting and use types more
appropriately.
(_mm_sll_epi32): Likewise.
(_mm_sll_epi64): Likewise.
(_mm_srl_epi16): Likewise.
(_mm_srl_epi32): Likewise.
(_mm_srl_epi64): Likewise.

Co-Authored-By: Jinsong Ji <jji@us.ibm.com>
From-SVN: r265507

emmintrin.h (_mm_sll_epi64): Remove wrong cast.

2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
    Jinsong Ji <jji@us.ibm.com>

* config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
* config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
__vector __bool int.  Use vec_cmpgt in preference to deprecated
function vec_vcmpgtfp.
(_mm_max_ps): Likewise.

Co-Authored-By: Jinsong Ji <jji@us.ibm.com>
From-SVN: r265506

rl78.c (insn_ok_now): Always re-recognize the insn if returning false.

* config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
if returning false.

From-SVN: r265505

* parser.c (cp_parser_sizeof_operand): Remove redundant grokdeclarator.

From-SVN: r265503

extend.texi (aligned): Expand attribute description.

gcc/ChangeLog:

* doc/extend.texi (aligned): Expand attribute description.
(Alignment): Rename section.  Discuss function arguments.

From-SVN: r265498

Use signed char in a test (PR testsuite/87739).

2018-10-25  Martin Liska  <mliska@suse.cz>

PR testsuite/87739
* gcc.dg/tree-ssa/pr84436-5.c (foo): Use signed char.

From-SVN: r265495

Revert partially changes from r265454 (PR other/87735).

2018-10-25  Martin Liska  <mliska@suse.cz>

PR other/87735
* gcc.dg/tree-prof/time-profiler-1.c: Revert.
* gcc.dg/tree-prof/time-profiler-2.c: Likewise.
* gcc.dg/tree-prof/time-profiler-3.c: Likewise.
2018-10-25  Martin Liska  <mliska@suse.cz>

PR other/87735
* libgcov-profiler.c: Revert.

From-SVN: r265494

PR libstdc++/87749 fix (and optimize) string move construction

The move constructor for the SSO string uses assign(const basic_string&)
when either:

(1) the source string is "local" and so the contents of the small string
buffer need to be copied, or

(2) the allocator does not propagate and is_always_equal is false.

Case (1) is suboptimal, because the assign member is not noexcept and
the compiler isn't smart enough to see it won't actually throw in this
case. This causes extra code in the move assignment operator so that any
exception will be turned into a call to std::terminate. This can be
fixed by copying small strings inline instead of calling assign.

Case (2) is a bug, because the specific instances of the allocators
could be equal even if is_always_equal is false. This can result in an
unnecessary deep copy (and potentially-throwing allocation) when the
storage should be moved. This can be fixed by simply checking if the
allocators are equal.

PR libstdc++/87749
* include/bits/basic_string.h [_GLIBCXX_USE_CXX11_ABI]
(basic_string::operator=(basic_string&&)): For short strings copy the
buffer inline. Only fall back to using assign(const basic_string&) to
do a deep copy when reallocation is needed.
* testsuite/21_strings/basic_string/modifiers/assign/char/87749.cc:
New test.
* testsuite/21_strings/basic_string/modifiers/assign/char/
move_assign_optim.cc: New test.
* testsuite/21_strings/basic_string/modifiers/assign/wchar_t/87749.cc:
New test.
* testsuite/21_strings/basic_string/modifiers/assign/wchar_t/
move_assign_optim.cc: New test.

From-SVN: r265493

ipa-devirt.c (main_odr_variant): Remove.

* ipa-devirt.c (main_odr_variant): Remove.
(hash_odr_name, types_same_for_odr, types_odr_comparable,
odr_name_hasher::equal, odr_subtypes_equivalent_p):
Drop use of main_odr_variant.
(add_type_duplicate): Silence confused warnings on integer types.
(get_odr_type): Always look for main variant.
(register_odr_type): Simplify.

From-SVN: r265492

tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Initialize ng to silence error with release checking bootstrap.

2018-10-25  Richard Biener  <rguenther@suse.de>

* tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
Initialize ng to silence error with release checking bootstrap.

From-SVN: r265491

S/390: Merge movdi_larl into movdi_64

Consider the following RTL:

(insn (set (mem/f/c:DI (reg/f:DI 60))
           (const:DI (plus:DI (symbol_ref:DI ("*.LANCHOR0"))
                              (const_int 8)))))

generated by cse2 pass.  It is matched to movdi_64, resulting in
the following inefficient code:

larl %r5,.L6 # Load literal pool@
lg %r1,.L7-.L6(%r5) # Load .LANCHOR0+8
stgrl %r1,.LANCHOR0
br %r14

Matching it to movdi_larl improves the code, eliminating one
instruction and the literal pool entry:

larl %r1,.LANCHOR0+8
stgrl %r1,.LANCHOR0
br %r14

Taking it one step further, there is no reason to keep movdi_64 and
movdi_larl separate, since this could potentially improve code in other
ways by giving lra one more alternative to choose from.

gcc/ChangeLog:

2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>

* config/s390/constraints.md (ZL): New constraint.
* config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
operands.
* config/s390/s390.md (movdi_larl): Remove.
(movdi_64): Add the LARL alternative.

gcc/testsuite/ChangeLog:

2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>

* gcc.target/s390/global-array-almost-huge-element.c: New test.
* gcc.target/s390/global-array-almost-negative-huge-element.c: New test.
* gcc.target/s390/global-array-element-pic.c: New test.
* gcc.target/s390/global-array-even-element.c: New test.
* gcc.target/s390/global-array-huge-element.c: New test.
* gcc.target/s390/global-array-negative-huge-element.c: New test.
* gcc.target/s390/global-array-odd-element.c: New test.

From-SVN: r265490

tree-if-conv.c: Include tree-ssa-sccvn.h.

2018-10-25  Richard Biener  <rguenther@suse.de>

* tree-if-conv.c: Include tree-ssa-sccvn.h.
(tree_if_conversion): Run CSE on the if-converted loop body.

From-SVN: r265489

Fix rtx_code_size static initialization order fiasco

r264556 and r264537 changed the format of EQ_ATTR_ALT RTXs to "ww",
which also required adjusting rtx_code_size initializer.  In order to
simplify things, the list of rtx_codes known to use HOST_WIDE_INTs was
replaced by the format string check.  However, unlike the old one, this
new check cannot be always performed at compile time, in which case a
static constructor is generated.  This may lead to a static
initialization order fiasco with respect to other static constructors
in the compiler, in case of PR87747, cselib's pool_allocator.

gcc/ChangeLog:

2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>

PR bootstrap/87747
* rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
(RTX_CODE_HWINT_P): New macro.
(rtx_code_size): Use RTX_CODE_HWINT_P ().

From-SVN: r265488

Relocation (= move+destroy)

2018-10-25  Marc Glisse  <marc.glisse@inria.fr>

PR libstdc++/87106
* include/bits/alloc_traits.h (_S_construct, _S_destroy, construct,
destroy): Add noexcept specification.
* include/bits/allocator.h (construct, destroy): Likewise.
* include/ext/alloc_traits.h (construct, destroy): Likewise.
* include/ext/malloc_allocator.h (construct, destroy): Likewise.
* include/ext/new_allocator.h (construct, destroy): Likewise.
* include/bits/stl_uninitialized.h (__relocate_object_a, __relocate_a,
__relocate_a_1): New functions.
(__is_trivially_relocatable): New class.
* include/bits/stl_vector.h (__use_relocate): New static member.
* include/bits/vector.tcc (reserve, _M_realloc_insert,
_M_default_append): Use __relocate_a.
(reserve, _M_assign_aux, _M_realloc_insert, _M_fill_insert,
_M_default_append, _M_range_insert): Move _GLIBCXX_ASAN_ANNOTATE_REINIT
after _Destroy.
* testsuite/23_containers/vector/modifiers/push_back/49836.cc:
Replace CopyConsOnlyType with DelAnyAssign.

From-SVN: r265485

ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types is anonymous.

* ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
is anonymous.
* g++.dg/lto/odr-1_0.C: New test.
* g++.dg/lto/odr-1_1.C: New test.

From-SVN: r265484

dg-cmp-results: display NA->FAIL & NA->UNRESOLVED by default

Currently, dg-cmp-results will not print anything for a test that was
not run before, even if it is a FAIL or UNRESOLVED now. This means that
when contributing a code change together with a testcase in the same
commit one must run dg-cmp-results twice: once to check for regression
on a full testsuite run and once against the new testcase with -v -v.
This also prevents using dg-cmp-results on sum files generated with
test_summary since these would not contain PASS.

This patch changes dg-cmp-results to print NA->FAIL and NA->UNRESOLVED
changes by default.

2018-10-25  Thomas Preud'homme  <thomas.preudhomme@linaro.org>

    contrib/
    * dg-cmp-results.sh: Print NA-FAIL and NA->UNRESOLVED changes at
    default verbosity.

From-SVN: r265483

[testsuite] Fix sibcall-9 & sibcall-10 with -fPIC

gcc.dg/sibcall-9.c and gcc.dg/sibcall-10.c give execution failure
on ARM when compiled with -fPIC due to the PIC access to volatile
variable v creating an extra spill which causes the frame size of the
two recursive functions to be different. Making the variable static
solve the issue because the variable can be access in a PC-relative way
and avoid the spill, while still testing sibling call as originally
intended.

2018-10-25  Thomas Preud'homme  <thomas.preudhomme@linaro.org>

gcc/testsuite/
    * gcc.dg/sibcall-9.c: Make v static.
    * gcc.dg/sibcall-10.c: Likewise.

From-SVN: r265482

re PR tree-optimization/87665 (gcc HEAD (svn: 265340) breaks elements on resize)

2018-10-25  Richard Biener  <rguenther@suse.de>

PR tree-optimization/87665
PR tree-optimization/87745
* tree-vectorizer.h (get_earlier_stmt): Remove.
(get_later_stmt): Pick up UID from the original non-pattern stmt.

* gfortran.dg/20181025-1.f: New testcase.

From-SVN: r265481

[DOC] Relocate list under Deprecated in options.texi to Var

gcc/doc
2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>

* options.texi (Deprecated): Move list to Var section.

From-SVN: r265480

re PR fortran/87725 (OpenMP 4.5 clause schedule(simd,monotonic:static) not understood)

PR fortran/87725
* openmp.c (gfc_match_omp_clauses): Parse simd, monotonic and
nonmonotonic modifiers regardless of if they have been parsed
already or if the opposite one has.  Fix up check whether
comma after modifier should be parsed.
(resolve_omp_clauses): Diagnose schedule modifier restrictions.

* c-c++-common/gomp/schedule-modifiers-1.c (bar): Separate modifier
from kind with a colon rather than comma.
* gfortran.dg/gomp/schedule-modifiers-1.f90: New test.
* gfortran.dg/gomp/schedule-modifiers-2.f90: New test.

From-SVN: r265479

Daily bump.

From-SVN: r265478

combine: Don't do make_more_copies for dest PC (PR87720)

Jumps are written in RTL as moves to PC.  But the latter has no mode,
so we shouldn't try to use it.  Since the optimization this routine
does does not really help for jumps at all, let's just skip it.

PR rtl-optimization/87720
* combine.c (make_more_copies): Skip if the dest is pc_rtx.

From-SVN: r265474

gOlogy: do not change code in isolate-paths for warnings only

The isolate-paths pass is activated by various -f flags, but also by
-Wnull-dereference.  Most of its codegen changes are conditioned on at
least one of the -f flags, but those that detect, warn about and
isolate paths that return the address of local variables are enabled
even if the pass is activated only by -Wnull-dereference.

-W flags should not cause codegen changes, so this patch makes the
codegen changes conditional on the presence of any of the -f flags
that activate the pass.  Should we have a separate option to activate
only this kind of transformation?

for  gcc/ChangeLog

* gimple-ssa-isolate-paths.c
(find_implicit_erroneous_behavior): Do not change code if the
pass is running for warnings only.
(find_explicit_erroneous_behavior): Likewise.

From-SVN: r265473

rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME): Define as rs6000_mangle_decl_assembler_name.

[gcc]
2018-10-24  Michael Meissner  <meissner@linux.ibm.com>

* config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
Define as rs6000_mangle_decl_assembler_name.
(rs6000_mangle_decl_assembler_name): If the user switched from IBM
long double to IEEE long double, switch the names of the long
double built-in functions to be <func>f128 instead of <func>l.

[gcc/testsuite]
2018-10-24  Michael Meissner  <meissner@linux.ibm.com>

* gcc.target/powerpc/float128-math.c: New test to make sure the
long double built-in function names use the f128 form if the user
switched from IBM long double to IEEE long double.
* gcc.target/powerpc/ppc-fortran/ieee128-math.f90: Likewise.

From-SVN: r265471

re PR c++/86288 (Recognize __gnu and/or __gnu__ as attribute-namespace)

PR c++/86288
* parser.c (cp_parser_std_attribute): Canonicalize attr_ns, and when
:: is not present and attr_ns non-NULL, canonicalize also attr_id.
(cp_parser_attribute_spec): Fix comment typo.

* g++.dg/cpp0x/gen-attrs-66.C: New test.

From-SVN: r265470

extend.texi (nonnull): List no-argument form.

gcc/ChangeLog:

* doc/extend.texi (nonnull): List no-argument form.  Reference
-fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.

From-SVN: r265469

PR c++/84851 - missing -Wclass-memaccess for a memcpy in a copy ctor with a non-trivial member

gcc/cp/ChangeLog:

PR c++/84851
* call.c (maybe_warn_class_memaccess): Tighten up.

gcc/testsuite/ChangeLog:

PR c++/84851
* g++.dg/Wclass-memaccess-4.C: Remove XFAIL.

From-SVN: r265467

tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.

2018-10-24  Richard Biener  <rguenther@suse.de>

* tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.

From-SVN: r265465

emmintrin.h (_mm_cvtpd_epi32): Change deprecated __vector long to __vector long long.

2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
    Jinsong Ji <jji@us.ibm.com>

* config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
__vector long to __vector long long.
(_mm_cvtpd_ps): Likewise.
(_mm_cvttpd_epi32): Likewise.
(_mm_cvtpi32_pd): Likewise.
(_mm_unpackhi_epi64): Likewise.
(_mm_unpacklo_epi64): Likewise.

From-SVN: r265464

Switch conversion: support any ax + b transformation (PR tree-optimization/84436).

2018-10-24  Martin Liska  <mliska@suse.cz>

PR tree-optimization/84436
* tree-switch-conversion.c (switch_conversion::contains_same_values_p):
Remove.
(switch_conversion::contains_linear_function_p): New.
(switch_conversion::build_one_array): Support linear
transformation on input.
* tree-switch-conversion.h (struct switch_conversion): Add
contains_linear_function_p declaration.
2018-10-24  Martin Liska  <mliska@suse.cz>

PR tree-optimization/84436
* gcc.dg/tree-ssa/pr84436-1.c: New test.
* gcc.dg/tree-ssa/pr84436-2.c: New test.
* gcc.dg/tree-ssa/pr84436-3.c: New test.
* gcc.dg/tree-ssa/pr84436-4.c: New test.
* gcc.dg/tree-ssa/pr84436-5.c: New test.

From-SVN: r265463

Return hash of ADDR_EXPR if its argument is CONSTANT_CLASS_P.

2018-10-24  Richard Biener  <rguenther@suse.de>

* varasm.c (const_hash_1): Return hash of ADDR_EXPR
if its argument is CONSTANT_CLASS_P.

From-SVN: r265462

ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL; it is wrong for forward declarations.

* ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
it is wrong for forward declarations.

From-SVN: r265460

Add myself to MAINTAINERS

ChangeLog:

2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>

* MAINTAINERS (Write After Approval): Add myself.

From-SVN: r265459