From d18ecd5b9e5dde58ae08a3eef1637406159ddaca Mon Sep 17 00:00:00 2001 From: YellowRoseCx <80486540+YellowRoseCx@users.noreply.github.com> Date: Thu, 10 Aug 2023 13:19:41 -0500 Subject: [PATCH] make mmq gen faster for amd --- ggml-cuda.cu | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/ggml-cuda.cu b/ggml-cuda.cu index fe29fc3c4df59..dc13cca3ffb66 100644 --- a/ggml-cuda.cu +++ b/ggml-cuda.cu @@ -72,7 +72,7 @@ #include "ggml.h" #define MIN_CC_DP4A 610 // minimum compute capability for __dp4a, an intrinsic for byte-wise dot products -#define CC_TURING 700 +#define CC_TURING 10000 #if defined(GGML_USE_HIPBLAS) #define __CUDA_ARCH__ 1300