Skip to content

feat: Configurable leading dimensions for Gemm functions with cuBLASLt #9

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

Open
sanjibansg wants to merge 6 commits into
base: main
Choose a base branch
from
Open

feat: Configurable leading dimensions for Gemm functions with cuBLASLt #9

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
aaf2d08
feat: configure leading dimensions explicitly
sanjibansg Mar 2, 2026
9aa4b88
fix: set stream while initializing handle
sanjibansg Mar 2, 2026
2498197
feat: print requested workspace
sanjibansg Mar 2, 2026
901c5ae
fix: provide layout order during adding
sanjibansg Mar 2, 2026
8a27ffe
fix: layout order configuration
sanjibansg Mar 3, 2026
54dce63
fix: remove look for requested workspace
sanjibansg Mar 3, 2026
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
17 changes: 7 additions & 10 deletions include/sofieBLAS/backends/cuda/sofieBLAS_cublas.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,7 +49,6 @@ struct PairEq {

class BlasCuda {
cublasLtHandle_t ltHandle = nullptr;
cublasHandle_t handle = nullptr;
cublasLtMatmulDesc_t operationDesc = nullptr;
cublasLtMatmulPreference_t preference = nullptr;
void *d_workspace = nullptr;
Expand All @@ -72,7 +71,6 @@ class BlasCuda {
BlasCuda(alpaka::QueueCudaRtNonBlocking &queue) : m_queue{queue} {
stream = static_cast<cudaStream_t>(m_queue.getNativeHandle());
CHECK_CUBLAS(cublasLtCreate(&ltHandle));
CHECK_CUBLAS(cublasCreate(&handle));
heuristic = {};
CHECK_CUBLAS(cublasLtMatmulDescCreate(&operationDesc, CUBLAS_COMPUTE_32F,
CUDA_R_32F));
Expand Down Expand Up @@ -118,10 +116,10 @@ class BlasCuda {
}
}

void AddLayoutConfig(std::size_t m, std::size_t n, std::size_t k) {
CheckAndAddLayout(k, m);
CheckAndAddLayout(k, n);
CheckAndAddLayout(m, n);
void AddLayoutConfig(std::size_t m, std::size_t n, std::size_t k, std::size_t lda, std::size_t ldb, std::size_t ldc) {
CheckAndAddLayout(k, m, lda);
CheckAndAddLayout(k, n, ldb);
CheckAndAddLayout(m, n, ldc);
}

template <typename T, typename TIdx>
Expand Down Expand Up @@ -171,7 +169,6 @@ gemm(char transa, char transb, const unsigned int m,
1,
&localHeuristic,
&returnedResults));

if (returnedResults == 0) {
cublasLtMatmulDescDestroy(localDesc);
std::cerr << "No suitable cuBLASLt algorithm found!\n";
Expand Down Expand Up @@ -238,7 +235,8 @@ gemmrelu(char transa, char transb, const unsigned int m,
1,
&localHeuristic,
&error_flag));

std::cout << "Requested workspace: "
<< localHeuristic.workspaceSize << std::endl;
if (error_flag == 0) {
cublasLtMatmulDescDestroy(localDesc);
std::cerr << "No suitable cuBLASLt algorithm found!\n";
Expand Down Expand Up @@ -313,11 +311,10 @@ gemmrelu(char transa, char transb, const unsigned int m,
private:
alpaka::QueueCudaRtNonBlocking m_queue;

void CheckAndAddLayout(size_t rows, size_t cols) {
void CheckAndAddLayout(size_t rows, size_t cols, size_t ld) {
auto key = std::make_pair(rows, cols);
if (LayoutStore.find(key) == LayoutStore.end()) {
cublasLtMatrixLayout_t temp = nullptr;
size_t ld = rows;
CHECK_CUBLAS(
cublasLtMatrixLayoutCreate(&temp, CUDA_R_32F, rows, cols, ld));
LayoutStore.emplace(key, temp);
Expand Down