Use syclcompat::dp4a

codeplaysoftware · Dec 5, 2024 · 174c0b9 · 174c0b9
1 parent e322028
commit 174c0b9
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/include/cutlass/numeric_conversion.h b/include/cutlass/numeric_conversion.h
@@ -3667,7 +3667,7 @@ struct NumericArrayConverter<float, int8_t, N, Round> {
 #if defined(__CUDA_ARCH__)
       t[ii] = __dp4a(x, mask[ii], 0);
 #else
-      t[ii] = x * mask[ii];
+      t[ii] = syclcompat::dp4a(x, mask[ii], 0);
 #endif
       r[ii] = static_cast<float>(t[ii]);
     }