Skip to content

Comm gemm fixes #2818

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
almogsegal wants to merge 2 commits into
base: main
Choose a base branch
from
+8 −3
Open

Comm gemm fixes #2818

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e32f3b8
Fix GemmRs B descriptor lld for transb=true
almogsegal Mar 31, 2026
263c33f
Set GemmRs communication type to output data type
almogsegal Mar 31, 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
11 changes: 8 additions & 3 deletions transformer_engine/common/comm_gemm/comm_gemm.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -186,9 +186,9 @@ void GemmRsInitMatrices(NVTECommGemmCtx* ctx, int64_t* ldd, int64_t m, int64_t n
}
if (transb) {
NVTE_CHECK(b1 == n, "Unsupported tensor dimension in B: expected ", n, ", got ", b1);
NVTE_CHECK_CUBLASMP(cublasMpMatrixDescriptorInit(
n, k, block_size(ctx, n), block_size(ctx, k), 0, 0, block_size(ctx, n),
get_cuda_dtype(b->dtype()), ctx->grid_row_major.get(), ctx->b_desc.get()));
NVTE_CHECK_CUBLASMP(cublasMpMatrixDescriptorInit(n, k, block_size(ctx, n), block_size(ctx, k),
0, 0, n, get_cuda_dtype(b->dtype()),
ctx->grid_row_major.get(), ctx->b_desc.get()));
} else {
NVTE_CHECK(b0 == n, "Unsupported tensor dimension in B: expected ", n, ", got ", b0);
NVTE_CHECK_CUBLASMP(cublasMpMatrixDescriptorInit(
Expand All @@ -200,6 +200,11 @@ void GemmRsInitMatrices(NVTECommGemmCtx* ctx, int64_t* ldd, int64_t m, int64_t n
NVTE_CHECK_CUBLASMP(cublasMpMatrixDescriptorInit(m, n, m, block_size(ctx, n), 0, 0, *ldd,
get_cuda_dtype(d->dtype()),
ctx->grid_row_major.get(), ctx->d_desc.get()));

const cudaDataType_t comm_type = get_cuda_dtype(d->dtype());
NVTE_CHECK_CUBLASMP(cublasMpMatmulDescriptorSetAttribute(
ctx->matmul_desc.get(), CUBLASMP_MATMUL_DESCRIPTOR_ATTRIBUTE_COMMUNICATION_TYPE, &comm_type,
sizeof comm_type));
}

void GemmArInitMatrices(NVTECommGemmCtx* ctx, int64_t* ldd, int64_t m, int64_t n, int64_t k,
Expand Down
Loading