vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llama : disable FA for AMD

Browse source
This commit is contained in:
Georgi Gerganov 2024-04-24 16:48:10 +03:00
parent 8937ec5307
commit ce281b904c
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
3 changed files with 12 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml-cuda/common.cuh
View file

@ -399,8 +399,8 @@ static __device__ __forceinline__ int __dp4a(const int a, const int b, int c) {
#define FP16_AVAILABLE defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__) ? \
defined(RDNA1) || defined(RDNA2) || defined(RDNA3) : __CUDA_ARCH__ >= CC_PASCAL
#define FP16_MMA_AVAILABLE defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__) ? \
defined(RDNA3) : __CUDA_ARCH__ >= CC_VOLTA
#define FP16_MMA_AVAILABLE !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_VOLTA
// TODO: move to ggml-common.h
static const __device__ int8_t kvalues_iq4nl[16] = {-127, -104, -83, -65, -49, -35, -22, -10, 1, 13, 25, 38, 53, 69, 89, 113};

3
ggml-cuda/fattn.cu
View file

@ -2,7 +2,10 @@
#include "fattn.cuh"
#include <cstdint>
#if FP16_MMA_AVAILABLE
#include <mma.h>
#endif
#define FATTN_KQ_STRIDE 256
#define HALF_MAX_HALF __float2half(65504.0f/2) // Use neg. of this instead of -INFINITY to initialize KQ max vals to avoid NaN upon subtraction.