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

Fix Cholesky priorities #999

Merged
merged 1 commit into from
Oct 2, 2023
Merged

Fix Cholesky priorities #999

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
18 changes: 8 additions & 10 deletions include/dlaf/factorization/cholesky/impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -139,7 +139,7 @@ void Cholesky<backend, device, T>::call_L(Matrix<T, device>& mat_a) {
// Cholesky decomposition on mat_a.readwrite(k,k) r/w potrf (lapack operation)
auto kk = LocalTileIndex{k, k};

potrfDiagTile<backend>(thread_priority::normal, mat_a.readwrite(kk));
potrfDiagTile<backend>(thread_priority::high, mat_a.readwrite(kk));

for (SizeType i = k + 1; i < nrtile; ++i) {
// Update panel mat_a.readwrite(i,k) with trsm (blas operation), using data mat_a.read(k,k)
Expand All @@ -160,7 +160,7 @@ void Cholesky<backend, device, T>::call_L(Matrix<T, device>& mat_a) {
for (SizeType i = j + 1; i < nrtile; ++i) {
// Update remaining trailing matrix mat_a.readwrite(i,j), reading
// mat_a.read(i,k) and mat_a.read(j,k), using gemm (blas operation)
gemmTrailingMatrixTile<backend>(thread_priority::normal, mat_a.read(LocalTileIndex{i, k}),
gemmTrailingMatrixTile<backend>(trailing_matrix_priority, mat_a.read(LocalTileIndex{i, k}),
mat_a.read(LocalTileIndex{j, k}),
mat_a.readwrite(LocalTileIndex{i, j}));
}
Expand Down Expand Up @@ -193,7 +193,7 @@ void Cholesky<backend, device, T>::call_L(comm::CommunicatorGrid grid, Matrix<T,

// Factorization of diagonal tile and broadcast it along the k-th column
if (kk_rank == this_rank)
potrfDiagTile<backend>(thread_priority::normal, mat_a.readwrite(kk_idx));
potrfDiagTile<backend>(thread_priority::high, mat_a.readwrite(kk_idx));

// If there is no trailing matrix
const SizeType kt = k + 1;
Expand Down Expand Up @@ -260,9 +260,7 @@ void Cholesky<backend, device, T>::call_L(comm::CommunicatorGrid grid, Matrix<T,
continue;

const auto i = distr.localTileFromGlobalTile<Coord::Row>(i_idx);
// TODO: This was using executor_np. Was that intentional, or should it
// be trailing_matrix_executor/priority?
gemmTrailingMatrixTile<backend>(thread_priority::normal, panel.read({Coord::Row, i}),
gemmTrailingMatrixTile<backend>(trailing_matrix_priority, panel.read({Coord::Row, i}),
panelT.read({Coord::Col, j}),
mat_a.readwrite(LocalTileIndex{i, j}));
}
Expand All @@ -285,7 +283,7 @@ void Cholesky<backend, device, T>::call_U(Matrix<T, device>& mat_a) {
for (SizeType k = 0; k < nrtile; ++k) {
auto kk = LocalTileIndex{k, k};

potrfDiagTile<backend>(thread_priority::normal, mat_a.readwrite(kk));
potrfDiagTile<backend>(thread_priority::high, mat_a.readwrite(kk));

for (SizeType j = k + 1; j < nrtile; ++j) {
trsmPanelTile<backend>(thread_priority::high, mat_a.read(kk),
Expand All @@ -300,7 +298,7 @@ void Cholesky<backend, device, T>::call_U(Matrix<T, device>& mat_a) {
mat_a.readwrite(LocalTileIndex{i, i}));

for (SizeType j = i + 1; j < nrtile; ++j) {
gemmTrailingMatrixTile<backend>(thread_priority::normal, mat_a.read(LocalTileIndex{k, i}),
gemmTrailingMatrixTile<backend>(trailing_matrix_priority, mat_a.read(LocalTileIndex{k, i}),
mat_a.read(LocalTileIndex{k, j}),
mat_a.readwrite(LocalTileIndex{i, j}));
}
Expand Down Expand Up @@ -333,7 +331,7 @@ void Cholesky<backend, device, T>::call_U(comm::CommunicatorGrid grid, Matrix<T,

// Factorization of diagonal tile and broadcast it along the k-th column
if (kk_rank == this_rank) {
potrfDiagTile<backend>(thread_priority::normal, mat_a.readwrite(kk_idx));
potrfDiagTile<backend>(thread_priority::high, mat_a.readwrite(kk_idx));
}

// If there is no trailing matrix
Expand Down Expand Up @@ -401,7 +399,7 @@ void Cholesky<backend, device, T>::call_U(comm::CommunicatorGrid grid, Matrix<T,

const auto j = distr.localTileFromGlobalTile<Coord::Col>(j_idx);

gemmTrailingMatrixTile<backend>(thread_priority::normal, panelT.read({Coord::Row, i}),
gemmTrailingMatrixTile<backend>(trailing_matrix_priority, panelT.read({Coord::Row, i}),
panel.read({Coord::Col, j}),
mat_a.readwrite(LocalTileIndex{i, j}));
}
Expand Down
Loading