Skip to content

HIP: fix RDNA4 build #17792

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

Merged
CISC merged 1 commit into from
Dec 5, 2025
+3 −4
Merged

HIP: fix RDNA4 build #17792

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
7 changes: 3 additions & 4 deletions ggml/src/ggml-cuda/mma.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -560,7 +560,7 @@ namespace ggml_cuda_mma {
xi[0] = xs[0];
xi[1] = xs[1];
#endif // defined(RDNA4)
}else if constexpr (I == 16 && J == 8) {
} else if constexpr (I == 16 && J == 8) {
int64_t * xi = (int64_t *) t.x;
#if defined(RDNA4)
const int64_t * xs = (int64_t *) ((const int *) xs0 + (threadIdx.x % t.I) * stride + 4 * (threadIdx.x / t.I));
Expand All @@ -577,14 +577,13 @@ namespace ggml_cuda_mma {
const int64_t * xs1 = xs + 2;
xi[2] = xs1[0];
xi[3] = xs1[1];

}else{
#endif // defined(RDNA4)
} else {
NO_DEVICE_CODE;
}
} else {
NO_DEVICE_CODE;
}
#endif // defined(RDNA4)
#else
#pragma unroll
for (int l = 0; l < t.ne; ++l) {
Expand Down
Loading