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

CUDA: fix MMQ writeback for int8 tensor cores #8100

Merged
merged 1 commit into from
Jun 24, 2024
Merged

CUDA: fix MMQ writeback for int8 tensor cores #8100

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
4 changes: 1 addition & 3 deletions ggml-cuda/mmq.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2054,15 +2054,13 @@ static __device__ __forceinline__ void mmq_write_back_mma(
static_assert(nwarps*mma_C::I == mmq_y, "nwarps*mma_C::I != mmq_y");
#endif // INT8_MMA_AVAILABLE

dst += (threadIdx.y % ntx) * mma_C::J*stride;

#pragma unroll
for (int j0 = 0; j0 < mmq_x; j0 += ntx*mma_C::J) {
#pragma unroll
for (int n = 0; n < ntx; ++n) {
#pragma unroll
for (int l = 0; l < mma_C::ne; ++l) {
const int j = j0 + mma_C::get_j(l);
const int j = j0 + (threadIdx.y % ntx) * mma_C::J + mma_C::get_j(l);

if (j > j_max) {
continue;
Expand Down
Loading