Ian Lance Taylor [Mon, 12 Jun 2017 23:14:05 +0000 (23:14 +0000)]
runtime: ignore _Gscan bit when checking status in CgocallDone
Also always access the atomicstatus field atomically.
The effect of not checking the _Gscan bit is that if the GC decides to
scan the stack just as the goroutine is leaving the system call, the
goroutine might fail to call exitsyscall. Then then typically causes
a runtime assertion failure later on. If we do call exitsyscall as we
should, it will stall (in casgstatus) until the _Gscan bit is cleared.
No separate test. I've observed causing sporadic failures running the
misc/cgo tests, but we don't currently have a way to run those
routinely for gccgo. I should fix that.
Pedro Alves [Mon, 12 Jun 2017 22:22:39 +0000 (22:22 +0000)]
Finish implementing P0426R1 "Constexpr for std::char_traits" for C++17
As discussed in PR c++/80265 ("__builtin_{memcmp,memchr,strlen} are
not usable in constexpr functions"), use __builtin_constant_p to tell
whether we can defer to a constexpr algorithm.
I used __always_inline__ just to be thorough. It isn't really really
necessary as far as I could determine.
* include/bits/char_traits.h (_GLIBCXX_ALWAYS_INLINE): Define if
not already defined.
(__cpp_lib_constexpr_char_traits): Uncomment.
(__constant_string_p, __constant_char_array_p): New.
(std::char_traits<char>, std::char_traits<wchar_t>): Add
_GLIBCXX17_CONSTEXPR on compare, length and find and use
__constant_string_p, __constant_char_array_p and
__builtin_constant_p to defer to __gnu_cxx::char_traits at compile
time.
Ian Lance Taylor [Mon, 12 Jun 2017 17:51:52 +0000 (17:51 +0000)]
compiler: avoid orphaning Bexpressions when processing conversions
The method Type_conversion_expression::do_get_backend was (in some
circumstances) creating a Bexpression for the source expression of the
conversion and then throwing it away before using it. Fix up this
method to insure that the call to get_backend() on the source
expression is only made when the result will be used.
Jonathan Wakely [Mon, 12 Jun 2017 16:37:28 +0000 (17:37 +0100)]
PR libstdc++/55917 do not handle exceptions in std::thread
PR libstdc++/55917
* src/c++11/thread.cc (execute_native_thread_routine): Remove
try-block so that exceptions propagate out of the thread and terminate
is called by the exception-handling runtime.
(execute_native_thread_routine_compat): Likewise.
* testsuite/30_threads/thread/cons/terminate.cc: New.
Martin Liska [Mon, 12 Jun 2017 11:56:32 +0000 (13:56 +0200)]
Create an extra BB in profile-generate (PR tree-optimization/81041).
2017-06-12 Martin Liska <mliska@suse.cz>
PR tree-optimization/81041
* tree-profile.c (gimple_gen_ic_func_profiler):
Create an extra BB in profile-generate
(gimple_gen_time_profiler): Likewise.
Richard Biener [Mon, 12 Jun 2017 10:42:57 +0000 (10:42 +0000)]
re PR tree-optimization/81053 (ICE on valid code at -O3 on x86_64-linux-gnu: in as_a, at is-a.h:192)
2017-06-12 Richard Biener <rguenther@suse.de>
PR tree-optimization/81053
* tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
with backedge value not defined in loop. Simplify def stmt
compute.
Ian Lance Taylor [Mon, 12 Jun 2017 03:25:04 +0000 (03:25 +0000)]
elf.c (backtrace_initialize): Always set *fileline_fn.
* elf.c (backtrace_initialize): Always set *fileline_fn.
* ttest.c: New file.
* btest.c: Move support functions into testlib.c. Change calls to
check to pass file name.
* testlib.c: New file, copied from (part of) btest.c.
* testlib.h: New file, declarations for testlib.c.
* edtest.c: Use testlib.h and testlib.c.
* configure.ac: Test for -pthread, set HAVE_PTHREAD conditional.
* Makefile.am (btest_SOURCES): Add testlib.c.
(edtest_SOURCES): Likewise.
(CHECK_PROGRAMS): Add ttest if HAVE_PTHREAD.
(ttest_SOURCES, ttest_CFLAGS, ttest_LDADD): Define.
* configure, Makefile.in: Rebuild.
Jason Merrill [Sun, 11 Jun 2017 05:29:14 +0000 (01:29 -0400)]
tree.h (id_equal): New.
* tree.h (id_equal): New.
* dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
instead of strcmp of IDENTIFIER_POINTER.
c-family/
* c-ada-spec.c, c-pragma.c: Use it.
cp/
* cp-tree.h, decl2.c, mangle.c, parser.c, pt.c, semantics.c: Use it.
Jan Hubicka [Sat, 10 Jun 2017 23:18:44 +0000 (01:18 +0200)]
ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
* ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
(mark_all_inlined_calls_cdtor): Fix formating.
(inline_transform): Rescale profile before inlining.
Jan Hubicka [Sat, 10 Jun 2017 20:37:34 +0000 (22:37 +0200)]
predict.c (drop_profile): Also drop individual bb/edge and cgraph edge counts.
* predict.c (drop_profile): Also drop individual bb/edge and cgraph
edge counts.
(handle_missing_profiles): Fix computation of tp_first_run.
(counts_to_freqs): Do not touch freqs when count is 0.
Tom de Vries [Sat, 10 Jun 2017 06:59:49 +0000 (06:59 +0000)]
Add effective target signal
2017-06-10 Tom de Vries <tom@codesourcery.com>
* lib/target-supports.exp (check_effective_target_signal): New proc.
* lib/gcc.exp (gcc_target_compile): Remove appending of
-DSIGNAL_SUPPRESS to additional_flags.
* gcc.c-torture/execute/ieee/ieee.exp: Add -DSIGNAL_SUPPRESS to
additional_flags if effective target signal is not supported by the
target.
* gcc.c-torture/execute/20101011-1.c: Same.
* gcc.dg/c99-stdint-1.c: Same.
* gcc.dg/c99-stdint-2.c: Same.
* gcc.dg/c99-stdint-5.c: Same.
* gcc.dg/c99-stdint-6.c: Same.
* gcc.dg/stdint-width-1.c: Same.
* doc/sourcebuild.texi (Effective-Target Keywords, Environment
attributes): Document signal effective target.
Jason Merrill [Sat, 10 Jun 2017 00:40:50 +0000 (20:40 -0400)]
Missing bits from N4268, constant evaluation for all non-type args.
* call.c (build_converted_constant_expr): Rename from
build_integral_nontype_arg_conv, handle all types.
* pt.c (convert_nontype_argument): In C++17 call it for all types.
Move NOP stripping inside pointer case, don't strip ADDR_EXPR.
* cvt.c (strip_fnptr_conv): Also strip conversions to the same type.
Jason Merrill [Sat, 10 Jun 2017 00:40:44 +0000 (20:40 -0400)]
Overhaul pointer-to-member conversion and template argument handling.
* call.c (standard_conversion): Avoid creating ck_pmem when the
class type is the same.
* cvt.c (can_convert_qual): Split from
perform_qualification_conversions.
* constexpr.c (cxx_eval_constant_expression): Check it.
* typeck.c (convert_ptrmem): Only cplus_expand_constant if
adjustment is necessary.
* pt.c (check_valid_ptrmem_cst_expr): Compare class types.
(convert_nontype_argument): Avoid redundant error.
David Malcolm [Fri, 9 Jun 2017 20:57:38 +0000 (20:57 +0000)]
Add support for mutually-incompatible fix-it hints
This patch adds a method:
rich_location::fixits_cannot_be_auto_applied
for ensuring that mutually-incompatible fix-its hints don't
lead to insane output from -fdiagnostics-generate-patch.
Fix-it hints within such rich_location instances are printed
as normal by diagnostic_show_locus, but don't affect the output
of -fdiagnostics-generate-patch.
gcc/ChangeLog:
* diagnostic.c (diagnostic_report_diagnostic): Only add fixits
to the edit_context if they can be auto-applied.
gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic-test-show-locus-bw.c
(test_mutually_exclusive_suggestions): New test function.
* gcc.dg/plugin/diagnostic-test-show-locus-generate-patch.c
(test_mutually_exclusive_suggestions): New test function.
* gcc.dg/plugin/diagnostic-test-show-locus-parseable-fixits.c
(test_mutually_exclusive_suggestions): New test function.
* gcc.dg/plugin/diagnostic_plugin_test_show_locus.c
(test_show_locus): Add special-case for
"test_mutually_exclusive_suggestions".
libcpp/ChangeLog:
* include/line-map.h
(rich_location::fixits_cannot_be_auto_applied): New method.
(rich_location::fixits_can_be_auto_applied_p): New accessor.
(rich_location::m_fixits_cannot_be_auto_applied): New field.
* line-map.c (rich_location::rich_location): Initialize new field.
opts.c (finish_options): If -fsplit-stack, disable implicit -forder-blocks-and-partition.
gcc/:
* opts.c (finish_options): If -fsplit-stack, disable implicit
-forder-blocks-and-partition.
* doc/invoke.texi (Optimize Options): Document that when using
-fsplit-stack -forder-blocks-and-partition is not implicitly
enabled.
gcc/go/:
* go-lang.c (go_langhook_post_options): If -fsplit-stack is turned
on, disable implicit -forder-blocks-and-partition.
gcc/testsuite/:
* gcc.dg/tree-prof/split-1.c: New test.
* class.c (build_vtbl_initializer): Mark dvirt_fn as cold.
* decl.c (cxx_init_decl_processing, push_throw_library_fn): Likewise.
(excpet.c): Mark terminate as cold.
* gcc.dg/predict-14.c: Avoid cold function detection.
* gcc.target/i386/umod-3.c: Likewise.
Tom de Vries [Fri, 9 Jun 2017 13:11:52 +0000 (13:11 +0000)]
Add "sorry, target cannot support label values" for nvptx
2017-06-09 Tom de Vries <tom@codesourcery.com>
PR target/80855
* config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
"target cannot support label values" when encountering LABEL_REF.
2017-06-09 Thomas Preud'homme <thomas.preudhomme@arm.com>
gcc/testsuite/
* lib/target-supports.exp (check_effective_target_arm_arch_FUNC_ok):
Test for null definitions instead of them being undefined. Add entry
for ARMv7VE. Reindent entry for ARMv8-M Baseline. Add comment warning
about using the effective target for architecture extension.
(check_effective_target_arm_arch_v7ve_ok): Remove.
(add_options_for_arm_arch_v7ve): Likewise.
Martin Liska [Fri, 9 Jun 2017 12:54:53 +0000 (14:54 +0200)]
Speed-up indirect-call instrumentation
2017-06-09 Martin Liska <mliska@suse.cz>
* tree-profile.c (gimple_gen_ic_profiler): Update comment.
(gimple_gen_ic_func_profiler): Emit direct comparison
of __gcov_indirect_call_callee with NULL.
(gimple_gen_time_profiler): Change probability from
PROB_VERY_UNLIKELY to PROB_UNLIKELY.
2017-06-09 Martin Liska <mliska@suse.cz>
* libgcov-profiler.c (__gcov_indirect_call_profiler_v2):
Reset __gcov_indirect_call_callee to NULL.
Richard Biener [Fri, 9 Jun 2017 09:40:45 +0000 (09:40 +0000)]
re PR tree-optimization/66623 (Unsafe FP math reduction used in strict math mode)
2017-06-09 Richard Biener <rguenther@suse.de>
PR tree-optimization/66623
* tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
refactor check_reduction into two parts, properly computing
whether we have to check reduction validity for outer loop
vectorization.
Richard Biener [Fri, 9 Jun 2017 09:36:06 +0000 (09:36 +0000)]
re PR tree-optimization/79483 ([graphite] ICE: verify_ssa failed (error: definition in block 31 does not dominate use in block 28))
2017-06-09 Richard Biener <rguenther@suse.de>
PR tree-optimization/79483
* graphite-scop-detection.c (order): New global.
(get_order): Compute bb to order mapping that satisfies code
generation constraints.
(cmp_pbbs): New helper.
(build_scops): Start domwalk at entry block, sort generated
pbbs.
Martin Liska [Fri, 9 Jun 2017 07:38:52 +0000 (09:38 +0200)]
Come up with selftests for predict.c.
2017-06-09 Martin Liska <mliska@suse.cz>
* predict.c (struct branch_predictor): New struct.
(test_prediction_value_range): New test.
(predict_c_tests): New function.
* selftest-run-tests.c (selftest::run_tests): Run the function.
* selftest.h: Declare new tests.
Fix lvalue/rvalue context mixup Set_and_use_temporary_expression's
do_get_backend() method. Enhance Mark_lvalue_varexprs to handle
conversions and temporary reference expressions, since occasionally
the front end emits code such as "deref(conv(tempref)) = ...".
If there is a large stack frame the rs6000 -fstack-limit code would
calculate the new stack pointer value using two insns (an addis and
an addi), with r0 as temporary. Such instructions do not exist.
This patch changes add<mode>3 to expand using a different strategy in
such cases; to FAIL if there is no way to do it (namely, if the source
is r0 and there is no way to get a temporary reg); and it changes
rs6000_emit_allocate_stack to assert gen_add3_insn did in fact emit
instructions.
PR target/80966
* config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
gen_add3_insn did not fail.
* config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
r0, construct that number in a temporary reg and add that reg to r0.
If asked to put the result in r0 as well, fail.
gcc/testsuite/
* gcc.target/powerpc/stack-limit.c: New testcase.
Jakub Jelinek [Thu, 8 Jun 2017 18:55:04 +0000 (20:55 +0200)]
re PR c++/81011 (ICE with #pragma omp task and inaccessible copy-constructor)
PR c++/81011
* cp-gimplify.c (cxx_omp_finish_clause): When changing clause
to OMP_CLAUSE_SHARED, also clear OMP_CLAUSE_SHARED_FIRSTPRIVATE
and OMP_CLAUSE_SHARED_READONLY flags.
compiler: lvalue context fixes in Unary_expression::do_get_backend
A couple of the Backend::var_expression invocations in
Unary_expression::do_get_backend were selecting "lvalue" context
incorrectly (these var exprs were not in an assignment or LHS
position); fix is to change back to "rvalue" context.
compiler: fix undefined symbol error with unexported method
When an interface I1 in an imported package has an unexported method,
and is then embedded into another interface I2, in a different
package, that has other methods, and a type T2 is converted to I2, we
failed to ever define the required interface method table. Naturally
T2 must implement the unexported method, and must therefore either be
defined in the same package as I1, or embed a type from that package.
In this case the compiler was assuming that that package would define
the interface method table, but of course, since I2 was not defined in
that package, that did not happen.
The fix is to only assume that the interface method table will be
defined elsewhere in the case where T2 and I2 are defined in the same
package. The compiler ensures that all such interface method tables
are created, in Gogo::build_interface_method_tables. This requires
knowing the package in which an interface type is defined, a simple
tweak to the importer.
Testing this revealed that the special case for stub methods created
for the embedded unexported methods of T2 needs to be done for
function declarations as it currently is for function definitions, so
that the newly created interface method tables use the correct name.
Testing that revealed that the code to determine the pkgpath symbol
for such stub methods was wrong. It assumed that one could call
pkgpath_for_symbol on the pkgpath to get the pkgpath symbol. Would
that it twere so simple. Instead, add a function to look up the
package, which must be known, and fetch the pkgpath symbol.
Olivier Hainque [Thu, 8 Jun 2017 14:08:56 +0000 (14:08 +0000)]
t-vxworks (LIBGCC2_INCLUDES): Add path to wrn/coreip to the set of -I options...
2017-06-08 Olivier Hainque <hainque@adacore.com>
* config/t-vxworks (LIBGCC2_INCLUDES): Add path to wrn/coreip to
the set of -I options, support for direct inclusions of net/uio.h
by VxWorks header files via ioLib.h.
tree-chkp.c (chkp_get_hard_register_var_fake_base_address): New function.
gcc/ChangeLog:
2017-05-09 Alexander Ivchenko <aivchenk@gmail.com>
* tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
New function.
(chkp_get_hard_register_fake_addr_expr): Ditto.
(chkp_build_addr_expr): Add check for hard reg case.
(chkp_parse_array_and_component_ref): Ditto.
(chkp_find_bounds_1): Ditto.
(chkp_process_stmt): Don't generate bounds store for
hard reg case.
gcc/testsuite/ChangeLog:
2017-05-09 Alexander Ivchenko <aivchenk@gmail.com>
* gcc.target/i386/mpx/hard-reg-2-lbv.c: New test.
* gcc.target/i386/mpx/hard-reg-2-nov.c: New test.
* gcc.target/i386/mpx/hard-reg-2-ubv.c: New test.
Jan Hubicka [Thu, 8 Jun 2017 13:16:44 +0000 (15:16 +0200)]
predict.c (maybe_hot_bb_p): Do not check profile status.
* predict.c (maybe_hot_bb_p): Do not check profile status.
(maybe_hot_edge_p): Likewise.
(probably_never_executed): Check for zero counts even if profile
is not read.
(unlikely_executed_edge_p): New function.
(unlikely_executed_stmt_p): New function.
(unlikely_executed_bb_p): New function.
(set_even_probabilities): Use unlikely predicates.
(combine_predictions_for_bb): Likewise.
(predict_paths_for_bb): Likewise.
(predict_paths_leading_to_edge): Likewise.
(determine_unlikely_bbs): New function.
(estimate_bb_frequencies): Use it.
(compute_function_frequency): Use zero counts even if profile is
not read.
* profile-count.h: Fix typo.
* g++.dg/tree-ssa/counts-1.C: New testcase.
* gcc.dg/tree-ssa/counts-1.c: New testcase.
projects / thirdparty / gcc.git / log
