tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-11 11:05:39 -04:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: Improve flash decoding kernel GPU occupancy for BS=1 case (#12183)

Browse Source 
- Find out active blocks per SM using cudaOccupancyMaxActiveBlocksPerMultiprocessor API. Use this value to determine the optimal parallel_blocks value.
- Prefer vector flash attention kernels over MMA kernel for BS=1

Fixes Issue: #12182
---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit is contained in:
Gaurav Garg
2025-03-20 01:22:06 +05:30
committed by GitHub
parent a9b59288e2
commit 517b5ddbf0
12 changed files with 214 additions and 266 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@@ -3230,6 +3230,9 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g
#ifndef FLASH_ATTN_AVAILABLE
return false;
#endif // FLASH_ATTN_AVAILABLE
if (op->src[0]->ne[3] != 1) {
return false;
}
if (op->src[1]->type == GGML_TYPE_BF16 || op->src[2]->type == GGML_TYPE_BF16) {
return false;
}