Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[CUSOLVER] workaround for issue #216 #217

Closed
wants to merge 1 commit into from
Closed

[CUSOLVER] workaround for issue #216 #217

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions src/lapack/backends/cusolver/cusolver_task.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,7 @@ static inline void host_task_internal(H &cgh, sycl::queue queue, F f) {
template <typename H, typename F>
static inline void onemkl_cusolver_host_task(H &cgh, sycl::queue queue, F f) {
(void)host_task_internal(cgh, queue, f);
queue.wait(); //temporary workaround for issue #216
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Unfortunately this does not work because it is actually syncing the queue before the command group associated with this onemkl_cusolver_host_task is submitted: The result is that I find this does not solve #216

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Interestingly I find that replacing queue.wait() with a change to host_task_internal:

static inline void host_task_internal(H &cgh, sycl::queue queue, F f) {
    cgh.interop_task([f, queue](sycl::interop_handler ih) {
        auto sc = CusolverScopedContextHandler(queue, ih);
        f(sc);
        auto handle = sc.get_handle(queue);
        cusolverStatus_t err;
        cudaStream_t currentStreamId;
        CUSOLVER_ERROR_FUNC(cusolverDnGetStream, err, handle, &currentStreamId);
        cuStreamSynchronize(currentStreamId);
    });
}

is also not solving #216. It seems that you have to either call cuStreamSynchronize from within the user facing onemkl_cusolver_host_task or call queue.wait() after the command group is submitted.

}

} // namespace cusolver
Expand Down
4 changes: 4 additions & 0 deletions tests/unit_tests/lapack/include/lapack_gtest_suite.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,23 +140,27 @@ using ComplexDoublePrecisionUsm = std::complex<double>;

#define DEFINE_TEST_DEPENDENCY_REAL(SUITE) \
TEST_P(SUITE##DependencyUsm, RealSinglePrecision) { \
GTEST_SKIP(); \
test_log::padding = "[ ] "; \
EXPECT_TRUE( \
dependency_controller.run(::usm_dependency<RealSinglePrecisionUsm>, *GetParam())); \
} \
TEST_P(SUITE##DependencyUsm, RealDoublePrecision) { \
GTEST_SKIP(); \
test_log::padding = "[ ] "; \
EXPECT_TRUE( \
dependency_controller.run(::usm_dependency<RealDoublePrecisionUsm>, *GetParam())); \
}

#define DEFINE_TEST_DEPENDENCY_COMPLEX(SUITE) \
TEST_P(SUITE##DependencyUsm, ComplexSinglePrecision) { \
GTEST_SKIP(); \
test_log::padding = "[ ] "; \
EXPECT_TRUE( \
dependency_controller.run(::usm_dependency<ComplexSinglePrecisionUsm>, *GetParam())); \
} \
TEST_P(SUITE##DependencyUsm, ComplexDoublePrecision) { \
GTEST_SKIP(); \
test_log::padding = "[ ] "; \
EXPECT_TRUE( \
dependency_controller.run(::usm_dependency<ComplexDoublePrecisionUsm>, *GetParam())); \
Expand Down