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

Jump to bottom

Check that SingleThreadedBlasScope is created and destroyed on the same std::thread #889

Merged
merged 3 commits into from
Jun 5, 2023
Merged

Check that SingleThreadedBlasScope is created and destroyed on the same std::thread #889

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9f546fc
Check that SingleThreadedBlasScope is created and destroyed on the sa…
msimberg May 30, 2023
6daae56
Narrow scope of SingleThreadedBlasScope in test_bt_band_to_tridiag to…
msimberg Jun 2, 2023
1e98e25
Narrow scope of SingleThreadedBlasScope in test_tridiag_solver_rot to…
msimberg Jun 5, 2023
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
7 changes: 7 additions & 0 deletions include/dlaf/common/single_threaded_blas.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,10 @@

#pragma once

#ifdef DLAF_ASSERT_MODERATE_ENABLE
#include <thread>
#endif

namespace dlaf::common::internal {
class [[nodiscard]] SingleThreadedBlasScope {
public:
Expand All @@ -21,6 +25,9 @@ class [[nodiscard]] SingleThreadedBlasScope {
SingleThreadedBlasScope& operator=(SingleThreadedBlasScope const&) = delete;

private:
#ifdef DLAF_ASSERT_MODERATE_ENABLE
std::thread::id calling_thread;
#endif
#ifdef DLAF_WITH_OPENMP
int omp_num_threads;
#endif
Expand Down
11 changes: 11 additions & 0 deletions src/common/single_threaded_blas.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,8 +17,15 @@

#include <dlaf/common/single_threaded_blas.h>

#ifdef DLAF_ASSERT_MODERATE_ENABLE
#include <dlaf/common/assert.h>
#endif

namespace dlaf::common::internal {
SingleThreadedBlasScope::SingleThreadedBlasScope() {
#ifdef DLAF_ASSERT_MODERATE_ENABLE
calling_thread = std::this_thread::get_id();
#endif
#ifdef DLAF_WITH_OPENMP
omp_num_threads = omp_get_max_threads();
omp_set_num_threads(1);
Expand All @@ -29,6 +36,10 @@ SingleThreadedBlasScope::SingleThreadedBlasScope() {
}

SingleThreadedBlasScope::~SingleThreadedBlasScope() {
#ifdef DLAF_ASSERT_MODERATE_ENABLE
auto current_thread = std::this_thread::get_id();
DLAF_ASSERT_MODERATE(calling_thread == current_thread, calling_thread, current_thread);
#endif
#ifdef DLAF_WITH_OPENMP
omp_set_num_threads(omp_num_threads);
#endif
Expand Down
60 changes: 32 additions & 28 deletions test/unit/eigensolver/test_bt_band_to_tridiag.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -134,24 +134,26 @@ void testBacktransformation(SizeType m, SizeType n, SizeType mb, SizeType nb, co
if (m == 0 || n == 0)
return;

dlaf::common::internal::SingleThreadedBlasScope single;
{
dlaf::common::internal::SingleThreadedBlasScope single;

using eigensolver::internal::nrStepsForSweep;
using eigensolver::internal::nrSweeps;
for (SizeType sweep = nrSweeps<T>(m) - 1; sweep >= 0; --sweep) {
for (SizeType step = nrStepsForSweep(sweep, m, b) - 1; step >= 0; --step) {
const SizeType j = sweep;
const SizeType i = j + 1 + step * b;
using eigensolver::internal::nrStepsForSweep;
using eigensolver::internal::nrSweeps;
for (SizeType sweep = nrSweeps<T>(m) - 1; sweep >= 0; --sweep) {
for (SizeType step = nrStepsForSweep(sweep, m, b) - 1; step >= 0; --step) {
const SizeType j = sweep;
const SizeType i = j + 1 + step * b;

const SizeType size = std::min(b, m - i);
const SizeType i_v = (i - 1) / b * b;
const SizeType size = std::min(b, m - i);
const SizeType i_v = (i - 1) / b * b;

T& v_head = *mat_hh_local.ptr({i_v, j});
const T tau = v_head;
v_head = 1;
T& v_head = *mat_hh_local.ptr({i_v, j});
const T tau = v_head;
v_head = 1;

using blas::Side;
lapack::larf(Side::Left, size, n, &v_head, 1, tau, mat_e_local.ptr({i, 0}), mat_e_local.ld());
using blas::Side;
lapack::larf(Side::Left, size, n, &v_head, 1, tau, mat_e_local.ptr({i, 0}), mat_e_local.ld());
}
}
}

Expand Down Expand Up @@ -212,24 +214,26 @@ void testBacktransformation(comm::CommunicatorGrid grid, SizeType m, SizeType n,
if (m == 0 || n == 0)
return;

dlaf::common::internal::SingleThreadedBlasScope single;
{
dlaf::common::internal::SingleThreadedBlasScope single;

using eigensolver::internal::nrStepsForSweep;
using eigensolver::internal::nrSweeps;
for (SizeType sweep = nrSweeps<T>(m) - 1; sweep >= 0; --sweep) {
for (SizeType step = nrStepsForSweep(sweep, m, b) - 1; step >= 0; --step) {
const SizeType j = sweep;
const SizeType i = j + 1 + step * b;
using eigensolver::internal::nrStepsForSweep;
using eigensolver::internal::nrSweeps;
for (SizeType sweep = nrSweeps<T>(m) - 1; sweep >= 0; --sweep) {
for (SizeType step = nrStepsForSweep(sweep, m, b) - 1; step >= 0; --step) {
const SizeType j = sweep;
const SizeType i = j + 1 + step * b;

const SizeType size = std::min(b, m - i);
const SizeType i_v = (i - 1) / b * b;
const SizeType size = std::min(b, m - i);
const SizeType i_v = (i - 1) / b * b;

T& v_head = *mat_hh_local.ptr({i_v, j});
const T tau = v_head;
v_head = 1;
T& v_head = *mat_hh_local.ptr({i_v, j});
const T tau = v_head;
v_head = 1;

using blas::Side;
lapack::larf(Side::Left, size, n, &v_head, 1, tau, mat_e_local.ptr({i, 0}), mat_e_local.ld());
using blas::Side;
lapack::larf(Side::Left, size, n, &v_head, 1, tau, mat_e_local.ptr({i, 0}), mat_e_local.ld());
}
}
}

Expand Down
12 changes: 7 additions & 5 deletions test/unit/eigensolver/test_tridiag_solver_rot.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -108,12 +108,14 @@ void testApplyGivenRotations(comm::CommunicatorGrid grid, const SizeType m, cons
const SizeType n = std::min((idx_end) *mb, m) - idx_begin * mb;
const GlobalElementSize offset(idx_begin * mb, idx_begin * mb);

dlaf::common::internal::SingleThreadedBlasScope single;
{
dlaf::common::internal::SingleThreadedBlasScope single;

for (auto rot : rots) {
T* x = mat_loc.ptr(GlobalElementIndex{0, rot.i} + offset);
T* y = mat_loc.ptr(GlobalElementIndex{0, rot.j} + offset);
blas::rot(n, x, 1, y, 1, rot.c, rot.s);
for (auto rot : rots) {
T* x = mat_loc.ptr(GlobalElementIndex{0, rot.i} + offset);
T* y = mat_loc.ptr(GlobalElementIndex{0, rot.j} + offset);
blas::rot(n, x, 1, y, 1, rot.c, rot.s);
}
}

auto result = [&dist = mat_h.distribution(), &mat_local = mat_loc](const GlobalElementIndex& element) {
Expand Down