nicolas
/
ik_llama_opt
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix new CUDA FA on Touring (#413) 

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow 2025-05-12 15:09:33 +03:00 committed by GitHub
parent ceb8f513e4
commit be1d5c4b7e
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-cuda/fattn-new-mma.cu
View File

@ -273,11 +273,11 @@ struct fattn_mma_f16_config<576, 512> {
}
static constexpr __device__ int get_nbatch_V2_device(int ncols) {
#if __CUDA_ARCH__ == GML_CUDA_CC_TURING
#if __CUDA_ARCH__ == CC_TURING
return ncols <= 16 ? 64 : 128;
#else
return ncols <= 16 ? 256 : 128;
#endif // __CUDA_ARCH__ == GML_CUDA_CC_TURING
#endif // __CUDA_ARCH__ == CC_TURING
}
static int get_nbatch_combine_host(const int /*cc*/, const int /*ncols*/) {