CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-17 11:37:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: fix FA occupancy, optimize tile kernel (#15982)

Browse Source
This commit is contained in:
Johannes Gäßler
2025-09-17 15:32:42 +02:00
committed by GitHub
parent cd08fc3ecc
commit c959b676be
4 changed files with 361 additions and 253 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
ggml/src/ggml-cuda/common.cuh
View File

@@ -75,6 +75,8 @@
#define GGML_CUDA_CC_IS_RDNA4(cc) (cc >= GGML_CUDA_CC_RDNA4)
#define GGML_CUDA_CC_IS_GCN(cc) (cc > GGML_CUDA_CC_OFFSET_AMD && cc < GGML_CUDA_CC_CDNA1)
#define GGML_CUDA_CC_IS_CDNA(cc) (cc >= GGML_CUDA_CC_CDNA1 && cc < GGML_CUDA_CC_RDNA1)
#define GGML_CUDA_CC_IS_CDNA1(cc) (cc >= GGML_CUDA_CC_CDNA1 && cc < GGML_CUDA_CC_CDNA2)
#define GGML_CUDA_CC_IS_CDNA2(cc) (cc >= GGML_CUDA_CC_CDNA2 && cc < GGML_CUDA_CC_CDNA3)
#define GGML_CUDA_CC_IS_CDNA3(cc) (cc >= GGML_CUDA_CC_CDNA3 && cc < GGML_CUDA_CC_RDNA1)
// Moore Threads
@@ -325,6 +327,20 @@ static constexpr __device__ int ggml_cuda_get_physical_warp_size() {
#endif // defined(GGML_USE_HIP) && (defined(__GFX9__) || defined(__GFX8__))
}
// Maximum number of bytes that can be copied in a single instruction.
static constexpr __device__ int ggml_cuda_get_max_cpy_bytes() {
#ifdef GGML_USE_HIP
return 16;
#else
#if __CUDA_ARCH__ >= GGML_CUDA_CC_VOLTA
return 16;
#else
return 8;
#endif // __CUDA_ARCH__ >= GGML_CUDA_CC_VOLTA
#endif // GGML_USE_HIP
}
[[noreturn]]
static __device__ void no_device_code(
const char * file_name, const int line, const char * function_name, const int arch, const char * arch_list) {