When compiling an empty loop:
...
long long v1;
#pragma acc parallel num_gangs (640) num_workers(1) vector_length (128)
#pragma acc loop
for (v1 = 0; v1 < 20; v1 += 2)
;
...
the compiler emits two subsequent bar.syncs. This triggers some bug on my
quadro m1200 (I'm assuming in the ptxas/JIT compiler) that hangs the testcase.
This patch works around the bug by doing an optimization: we detect that this is
an empty loop (a forked immediately followed by a joining), and don't emit the
barriers.
The patch does not include the test-case yet, since vector_length (128) is not
yet supported at this point.
2019-01-07 Tom de Vries <tdevries@suse.de>
PR target/85381
* config/nvptx/nvptx.c (nvptx_process_pars): Don't emit barriers for
empty loops.
From-SVN: r267630
Add support for a per-worker broadcast buffer and barrier, to be used for
openacc vector_length larger than warp-size.
2019-01-07 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (oacc_bcast_partition): Declare.
(nvptx_option_override): Init oacc_bcast_partition.
(nvptx_init_oacc_workers): New function.
(nvptx_declare_function_name): Call nvptx_init_oacc_workers.
(nvptx_needs_shared_bcast): New function.
(nvptx_find_par): Generalize to enable vectors to use shared-memory
to propagate state.
(nvptx_shared_propagate): Initialize vector bcast partition and
synchronization state.
(nvptx_single): Generalize to enable vectors to use shared-memory
to propagate state.
(nvptx_process_pars): Likewise.
(nvptx_set_current_function): Initialize oacc_broadcast_partition.
* config/nvptx/nvptx.h (struct machine_function): Add
bcast_partition and sync_bar members.
From-SVN: r267629
Allow PTX_MAX_VECTOR_LENGTH to be define as larger than PTX_WARP_SIZE in
nvptx_goacc_validate_dims_1.
2019-01-07 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (nvptx_welformed_vector_length_p)
(nvptx_apply_dim_limits): New function.
(nvptx_goacc_validate_dims_1): Allow PTX_MAX_VECTOR_LENGTH larger than
PTX_WARP_SIZE.
From-SVN: r267628
Move warnings in nvptx_goacc_validate_dims_1 to as late as possible. This
allows us more flexibility in setting the dimensions.
2019-01-07 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (nvptx_goacc_validate_dims_1): Move warnings to
as late as possible.
From-SVN: r267627
Remove PTX_VECTOR_LENGTH and replace uses of it with PTX_DEFAULT_VECTOR_LENGTH,
PTX_MAX_VECTOR_LENGTH and PTX_WARP_SIZE.
2019-01-07 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (PTX_VECTOR_LENGTH): Remove.
(PTX_DEFAULT_VECTOR_LENGTH, PTX_MAX_VECTOR_LENGTH): Define.
(nvptx_goacc_validate_dims_1, nvptx_dim_limit)
(nvptx_goacc_reduction_fini): Use PTX_DEFAULT_VECTOR_LENGTH,
PTX_MAX_VECTOR_LENGTH and PTX_WARP_SIZE instead of PTX_VECTOR_LENGTH.
From-SVN: r267626
Add a few asserts to nvptx_goacc_validate_dims.
2019-01-07 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Add asserts.
From-SVN: r267625
The vector-length-128-3.c test-case uses GOMP_OPENACC_DIM=-:-:128, but '-' is
not yet supported on trunk. Use GOMP_OPENACC_DIM=::128 instead.
2019-01-07 Tom de Vries <tdevries@suse.de>
* testsuite/libgomp.oacc-c-c++-common/vector-length-128-3.c: Fix
GOMP_OPENACC_DIM argument.
From-SVN: r267624
Expose oacc_min_dims to backends.
2019-01-07 Tom de Vries <tdevries@suse.de>
* omp-offload.c (oacc_get_min_dim): New function.
* omp-offload.h (oacc_get_min_dim): Declare.
From-SVN: r267623
PR target/88521
* config/i386/i386.c (function_value_ms_64): Return small sturct in
AX_REG and float/double in FIRST_SSE_REG for 4 or 8 byte modes.
From-SVN: r267622
At some point when the GFORTRAN_9 node was added it was forgotten to
make it depend on GFORTRAN_8. This patch fixes this.
Committed as obvious.
2019-01-07 Janne Blomqvist <jb@gcc.gnu.org>
* gfortran.map (GFORTRAN_9): Make GFORTRAN_9 node depend on
GFORTRAN_8.
From-SVN: r267621
Move std::filesystem directory iterators and operations from
libstdc++fs.a to main libstdc++ library. These components have many
dependencies on OS support, which is not available on all targets. Some
additional autoconf checks and conditional compilation is needed to
ensure the files will build for all targets. Previously this code was
not compiled without --enable-libstdcxx-filesystem-ts but the C++17
components should be available for all hosted builds.
The tests for these components no longer need to link to libstdc++fs.a,
but are not expected to pass on all targets. To avoid numerous failures
on targets which are not expected to pass the tests (due to missing OS
functionality) leave the dg-require-filesystem-ts directives in place
for now. This will ensure the tests only run for builds where the
filesystem-ts library is built, which presumably means some level of OS
support is present.
PR libstdc++/86756
* acinclude.m4 (GLIBCXX_CHECK_FILESYSTEM_DEPS): Check for utime and
lstat and define _GLIBCXX_USE_UTIME and _GLIBCXX_USE_LSTAT.
* config.h.in: Regenerate.
* config/abi/pre/gnu.ver (GLIBCXX_3.4.26): Export symbols for
remaining std::filesystem types and functions.
* configure: Regenerate.
* src/c++17/Makefile.am: Add C++17 filesystem sources.
* src/c++17/Makefile.in: Regenerate.
* src/c++17/cow-fs_dir.cc: Move src/filesystem/cow-std-dir.cc to
here, and change name of included file.
* src/c++17/cow-fs_ops.cc: Move src/filesystem/cow-std-ops.cc to
here, and change name of included file.
* src/c++17/fs_dir.cc: Move src/filesystem/std-dir.cc to here. Change
path to dir-common.h.
* src/c++17/fs_ops.cc: Move src/filesystem/std-ops.cc to here. Change
path to ops-common.h. Disable -Wunused-parameter warnings.
(internal_file_clock): Define unconditionally.
[!_GLIBCXX_HAVE_SYS_STAT_H] (internal_file_clock::from_stat): Do not
define.
(do_copy_file, do_space): Move definitions to ops.common.h.
(copy, file_size, hard_link_count, last_write_time, space): Only
perform operation when _GLIBCXX_HAVE_SYS_STAT_H is defined, otherwise
report an error.
(last_write_time, read_symlink): Remove unused attributes from
parameters.
* src/filesystem/Makefile.am: Remove C++17 filesystem sources.
* src/filesystem/Makefile.in: Regenerate.
* src/filesystem/cow-std-dir.cc: Move to src/c++17/cow-fs_dir.cc.
* src/filesystem/cow-std-ops.cc: Move to src/c++17/cow-fs_ops.cc.
* src/filesystem/std-dir.cc: Move to src/c++17/fs_dir.cc.
* src/filesystem/std-ops.cc: Move to src/c++17/fs_ops.cc.
* src/filesystem/dir-common.h [!_GLIBCXX_HAVE_DIRENT_H]: Define
dummy types and functions instead of using #error.
* src/filesystem/dir.cc [!_GLIBCXX_HAVE_DIRENT_H]: Use #error.
* src/filesystem/ops-common.h [!_GLIBCXX_USE_LSTAT] (lstat): Define
in terms of stat.
[!_GLIBCXX_HAVE_UNISTD_H]: Define dummy types and functions.
(do_copy_file, do_space): Move definitions here from std-ops.cc.
* src/filesystem/ops.cc: Adjust calls to do_copy_file and do_space
to account for new namespace.
* testsuite/27_io/filesystem/directory_entry/86597.cc: Remove
-lstdc++fs from dg-options.
* testsuite/27_io/filesystem/directory_entry/lwg3171.cc: Likewise.
* testsuite/27_io/filesystem/file_status/1.cc: Likewise.
* testsuite/27_io/filesystem/filesystem_error/cons.cc: Likewise.
* testsuite/27_io/filesystem/filesystem_error/copy.cc: Likewise.
* testsuite/27_io/filesystem/iterators/directory_iterator.cc:
Likewise.
* testsuite/27_io/filesystem/iterators/pop.cc: Likewise.
* testsuite/27_io/filesystem/iterators/recursive_directory_iterator.cc:
Likewise.
* testsuite/27_io/filesystem/operations/absolute.cc: Likewise.
* testsuite/27_io/filesystem/operations/canonical.cc: Likewise.
* testsuite/27_io/filesystem/operations/copy.cc: Likewise.
* testsuite/27_io/filesystem/operations/copy_file.cc: Likewise.
* testsuite/27_io/filesystem/operations/create_directories.cc:
Likewise.
* testsuite/27_io/filesystem/operations/create_directory.cc: Likewise.
* testsuite/27_io/filesystem/operations/create_symlink.cc: Likewise.
* testsuite/27_io/filesystem/operations/current_path.cc: Likewise.
* testsuite/27_io/filesystem/operations/equivalent.cc: Likewise.
* testsuite/27_io/filesystem/operations/exists.cc: Likewise.
* testsuite/27_io/filesystem/operations/file_size.cc: Likewise.
* testsuite/27_io/filesystem/operations/is_empty.cc: Likewise.
* testsuite/27_io/filesystem/operations/last_write_time.cc: Likewise.
* testsuite/27_io/filesystem/operations/permissions.cc: Likewise.
* testsuite/27_io/filesystem/operations/proximate.cc: Likewise.
* testsuite/27_io/filesystem/operations/read_symlink.cc: Likewise.
* testsuite/27_io/filesystem/operations/relative.cc: Likewise.
* testsuite/27_io/filesystem/operations/remove.cc: Likewise.
* testsuite/27_io/filesystem/operations/remove_all.cc: Likewise.
* testsuite/27_io/filesystem/operations/space.cc: Likewise.
* testsuite/27_io/filesystem/operations/status.cc: Likewise.
* testsuite/27_io/filesystem/operations/symlink_status.cc: Likewise.
* testsuite/27_io/filesystem/operations/temp_directory_path.cc:
Likewise.
* testsuite/27_io/filesystem/operations/weakly_canonical.cc: Likewise.
From-SVN: r267616
The previous patch for PR 87431 assumed that initialing a scalar type
could not throw, but it can obtain its value via a conversion operator,
which could throw. This meant the variant could get into a valueless
state, but the valueless_by_exception() member function would always
return false.
This patch fixes it by changing the emplace members to have strong
exception safety when initializing a contained value of trivially
copyable type. The _M_valid() member gets a corresponding change to
always return true for trivially copyable types, not just scalar types.
Strong exception safety (i.e. never becoming valueless) is achieved by
only replacing the current contained value once any potentially throwing
operations have completed. If constructing the new contained value can
throw then a new std::variant object is constructed to hold it, and then
move-assigned to *this (which won't throw).
PR libstdc++/87431
* include/std/variant (_Variant_storage<true, _Types...>::_M_valid):
Check is_trivially_copyable instead of is_scalar.
(variant::emplace<N, Args>(Args&&...)): If construction of the new
contained value can throw and its type is trivially copyable then
construct into a temporary variant and move from it, to provide the
strong exception safety guarantee.
(variant::emplace<N, U, Args>(initializer_list<U>, Args&&...)):
Likewise.
* testsuite/20_util/variant/87431.cc: New test.
* testsuite/20_util/variant/run.cc: Adjust test so that throwing
conversion causes valueless state.
From-SVN: r267614
2019-01-06 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/88658
* gfortran.h: Add macro gfc_real_4_kind
* simplify.c (simplify_min_max): Special case for the types of
AMAX0, AMIN0, MAX1 and MIN1, which actually change the types of
their arguments.
2019-01-06 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/88658
* gfortran.dg/min_max_type_2.f90: New test.
From-SVN: r267609
This verifies that the <bits/extc++.h> header can be compiled with ASCII
as the input character set.
PR libstdc++/88607
* testsuite/17_intro/headers/c++1998/charset.cc: New test.
* testsuite/17_intro/headers/c++2011/charset.cc: New test.
* testsuite/17_intro/headers/c++2014/charset.cc: New test.
* testsuite/17_intro/headers/c++2017/charset.cc: New test.
* testsuite/17_intro/headers/c++2020/charset.cc: New test.
From-SVN: r267607
In C++17 the clock used for filesystem::file_time_type is unspecified,
allowing it to be chrono::system_clock. The C++2a draft requires it to
be a distinct type, with additional member functions to convert to/from
other clocks (either the system clock or UTC). In order to avoid an ABI
change later, this patch defines a new distinct type now, which will be
used for std::chrono::file_clock later.
* include/bits/fs_fwd.h (__file_clock): Define new clock.
(file_time_type): Redefine in terms of __file_clock.
* src/filesystem/ops-common.h (file_time): Add FIXME comment about
overflow.
* src/filesystem/std-ops.cc (is_set(perm_options, perm_options)): Give
internal linkage.
(internal_file_lock): New helper type for accessing __file_clock.
(do_copy_file): Use internal_file_lock to convert system time to
file_time_type.
(last_write_time(const path&, error_code&)): Likewise.
(last_write_time(const path&, file_time_type, error_code&)): Likewise.
From-SVN: r267602
2019-01-05 Janus Weil <janus@gcc.gnu.org>
PR fortran/88009
* class.c (gfc_find_derived_vtab): Mark the _final component as
artificial.
(find_intrinsic_vtab): Ditto. Also add an extra check to avoid
dereferencing a null pointer and adjust indentation.
* resolve.c (resolve_fl_variable): Add extra check to avoid
dereferencing a null pointer. Move variable declarations to local scope.
(resolve_fl_procedure): Add extra check to avoid dereferencing a null
pointer.
* symbol.c (check_conflict): Suppress errors for artificial symbols.
2019-01-05 Janus Weil <janus@gcc.gnu.org>
PR fortran/88009
* gfortran.dg/blockdata_10.f90: New test case.
From-SVN: r267598
PR middle-end/82564
PR target/88620
* expr.c (expand_assignment): For calls returning VLA structures
if to_rtx is not a MEM, force it into a stack temporary.
* gcc.dg/nested-func-12.c: New test.
* gcc.c-torture/compile/pr82564.c: New test.
From-SVN: r267595
PR debug/88635
* dwarf2out.c (const_ok_for_output_1): Reject MINUS that contains
SYMBOL_REF, CODE_LABEL or UNSPEC in subexpressions of second argument.
Reject PLUS that contains SYMBOL_REF, CODE_LABEL or UNSPEC in
subexpressions of both operands.
(mem_loc_descriptor): Handle UNSPEC if target hook acks it and all the
subrtxes are CONSTANT_P.
* config/i386/i386.c (ix86_const_not_ok_for_debug_p): Revert
2018-11-09 changes.
* gcc.dg/debug/dwarf2/pr88635.c: New test.
From-SVN: r267594
Traceback routines, e.g. callers and funcentry, may call
__go_get_backtrace_state. If a profiling signal arrives while we
are in the critical section of __go_get_backtrace_state, it tries
to do a traceback, which also calls __go_get_backtrace_state,
which tries to enter the same critical section and will deadlock.
Prevent this deadlock by setting up runtime_in_callers before
calling __go_get_backtrace_state.
Found while investigating golang/go#29448. Will add a test in the
next CL.
Updates golang/go#29448.
Reviewed-on: https://go-review.googlesource.com/c/156037
From-SVN: r267590
The C++17 standard added some new members to std::basic_string, which
were not previously instantiated in the library. This meant that the
extern template declarations had to be disabled for C++17 mode. With
this patch the new members are instantiated in the library and so the
explicit instantiation declarations can be used for C++17.
The new members added by C++2a are still not exported, and so the
explicit instantiation declarations are still disabled for C++2a.
* config/abi/pre/gnu.ver (GLIBCXX_3.4.21): Make patterns less greedy
for const member functions of std::basic_string.
(GLIBCXX_3.4.26): Export member functions of std::basic_string added
in C++17.
* include/bits/basic_string.h (basic_string(__sv_wrapper, const A&)):
Make non-standard constructor private.
[!_GLIBCXX_USE_CXX11_ABI] (basic_string(__sv_wrapper, const A&)):
Likewise.
* include/bits/basic_string.tcc (std::string, std::wstring): Declare
explicit instantiations for C++17 as well as earlier dialects.
* src/c++17/Makefile.am: Add new source files.
* src/c++17/Makefile.in: Regenerate.
* src/c++17/cow-string-inst.cc: New file defining explicit
instantiations for basic_string member functions added in C++17.
* src/c++17/string-inst.cc: Likewise.
From-SVN: r267585
Add these constructors from C++11 which were missing from the COW
basic_string. Additionally simplify the definitions of the
basic_string::reference and basic_string::const_reference types as
required by C++11.
This allows filesystem::path::string<Allocator>() to be simplified, so
that the same code is used for both basic_string implementations.
* config/abi/pre/gnu.ver (GLIBCXX_3.4.26): Export allocator-extended
copy/move constructors for old std::basic_string.
* include/bits/basic_string.h [!_GLIBCXX_USE_CXX11_ABI]
(basic_string::reference, basic_string::const_reference): Define
as plain references for C++11 and later.
(basic_string::basic_string()): Put constructor body outside
preprocessor conditional groups.
(basic_string::basic_string(basic_string&&)): Move _Alloc_hider
instead of copying it.
(basic_string::basic_string(const basic_string&, const _Alloc&)):
Define.
(basic_string::basic_string(basic_string&&, const _Alloc&)):
Define.
* include/bits/fs_path.h [!_GLIBCXX_USE_CXX11_ABI]: Remove special
cases for old basic_string.
* testsuite/21_strings/basic_string/cons/char/8.cc: Test
allocator-extended constructors unconditionally. Add extra members to
allocator type when using old string ABI.
* testsuite/21_strings/basic_string/allocator/71964.cc: Enable test
for old string ABI.
* testsuite/21_strings/basic_string/cons/wchar_t/8.cc: Likewise.
From-SVN: r267584
This currently checks _GLIBCXX_USE_DUAL_ABI which is incorrect, as that
can be true when _GLIBCXX_USE_CXX11_ABI == 0. The correct check would be
_GLIBCXX_USE_CXX11_ABI == 1, but that's made redundant by the cxx11-abi
effective target that the test requires. However, the test will fail if
-fno-inline is used, so check __NO_INLINE__ instead.
* testsuite/23_containers/list/61347.cc: Avoid spurious failure when
-fno-inline added to test flags.
From-SVN: r267582
This test started passing with the old ABI with r263808, so doesn't need
to be marked XFAIL now.
* testsuite/21_strings/basic_string/requirements/
explicit_instantiation/debug.cc: Remove XFAIL for old ABI.
From-SVN: r267578
Using path::_List::erase(const_iterator) to remove a non-final component
in path::lexically_normal() is a bug, because it leaves the following
component with an incorrect _M_pos value.
Instead of providing erase members that allow removing components from
the middle, replace them with pop_back() and
_M_erase_from(const_iterator) which only allow removing elements at the
end. Most uses of erase are unaffected, because they only remove
elements from the end anyway. The one use of erasure from the middle in
lexically_normal() is replaced by calls to pop_back() and/or clearing
the last component to leave it as an empty final filename.
Also replace the "???" comment in lexically_normal() to document when
that branch is taken.
* include/bits/fs_path.h (path::_List::erase): Replace both overloads
with ...
(path::pop_back(), path::_M_erase_from(const_iterator)): New member
functions that will only erase elements at the end.
* src/filesystem/std-path.cc (path::_List::_Impl::pop_back()): Define.
(path::_List::_Impl::_M_erase_from(const_iterator)): Define.
(path::_List::operator=(const _List&)): Use _M_erase_from(p) instead
of erase(p, end()).
(path::_List::pop_back()): Define.
(path::_List::_M_erase_from(const_iterator)): Define.
(path::operator/=(const path&)): Use pop_back to remove last component
and _M_erase_from to remove multiple components.
(path::_M_append(basic_string_view<value_type>)): Likewise.
(path::operator+=(const path&)): Likewise.
(path::_M_concat(basic_string_view<value_type>)): Likewise.
(path::remove_filename()): Likewise.
(path::lexically_normal()): Use _List::_Impl iterators instead of
path::iterator. Use pop_back to remove components from the end. Clear
trailing filename, instead of using erase(const_iterator) to remove
a non-final component.
* testsuite/27_io/filesystem/path/generation/normal.cc: Test
additional cases.
* testsuite/27_io/filesystem/path/generation/normal2.cc: New test.
From-SVN: r267576
When erasing a trailing empty filename component, the output iterator
was not decremented, causing the next component to be created at the
wrong position.
* src/filesystem/std-path.cc (path::operator+=(const path&)): Fix
incorrect treatment of empty filename after trailing slash.
* testsuite/27_io/filesystem/path/concat/path.cc: Test problem case.
From-SVN: r267574
These tests rely on inlining, so if -fno-inline is added to the compiler
flags the tests fail. Use the predefined __NO_INLINE__ macro to detect
that situation, and don't bother testing the move assignment.
* testsuite/21_strings/basic_string/modifiers/assign/char/
move_assign_optim.cc: Avoid spurious failure when -fno-inline added
to test flags.
* testsuite/21_strings/basic_string/modifiers/assign/wchar_t/
move_assign_optim.cc: Likewise.
From-SVN: r267573
PR target/88594
* config/i386/i386.c (ix86_expand_divmod_libfunc): Use mode instead
of GET_MODE (opN) as modes of the libcall arguments.
* gcc.dg/pr88594.c: New test.
From-SVN: r267571
While their use for masking is indeed restricted to %k1...%k7, use as
"normal" insn operands also permits %k0. Remove the unnecessary
limitations, requiring quite a few testsuite adjustments.
Oddly enough some AVX512{F,DQ} test cases already check for %k[0-7],
while others did permit {%k0} - where they get touched here anyway this
gets fixed at the same time.
From-SVN: r267570
