fix AMD

ggerganov · Feb 10, 2024 · 2bb97fc · 2bb97fc
1 parent 97f8a7a
commit 2bb97fc
Showing 1 changed file with 5 additions and 5 deletions.
diff --git a/ggml-cuda.cu b/ggml-cuda.cu
@@ -5319,13 +5319,13 @@ static __global__ void mul_mat_vec_q(
     const void * __restrict__ vx, const void * __restrict__ vy, float * __restrict__ dst,
     const int ncols_x, const int nrows_x, const int nrows_y, const int nrows_dst) {
 
-#if __CUDA_ARCH__ < CC_RDNA2
-    constexpr int nwarps              = ncols_y <= 4 ? 4 : 2;
-    constexpr int rows_per_cuda_block = ncols_y == 1 ? 1 : 2;
-#else
+#if defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__) && (defined(RDNA2) || defined(RDNA3))
     constexpr int nwarps              = 1;
     constexpr int rows_per_cuda_block = 1;
-#endif // __CUDA_ARCH__ < CC_RDNA2
+#else
+    constexpr int nwarps              = ncols_y <= 4 ? 4 : 2;
+    constexpr int rows_per_cuda_block = ncols_y == 1 ? 1 : 2;
+#endif // defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__) && !defined(RDNA2) && !defined(RDNA3)
 
     constexpr int blocks_per_iter = vdr * nwarps*WARP_SIZE / qi;