ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-30 19:31:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

This seems very slightly better

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-09-02 18:24:45 +03:00
parent 32e223df46
commit 27e8ed6454
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cuda/fattn-mma-f16.cuh
View File

@@ -1400,7 +1400,7 @@ void launch_fattn_mma(
dim3 blocks_num; dim3 blocks_num;
if (stream_k) { if (stream_k) {
// For short contexts it can be faster to have the SMs work on whole tiles because this lets us skip the fixup. // For short contexts it can be faster to have the SMs work on whole tiles because this lets us skip the fixup.
const int max_blocks = 2*nsm; const int max_blocks = Q->ne[1] > 1 ? 2*nsm : nsm;
const int tiles_nwaves = (ntiles_total + max_blocks - 1) / max_blocks; const int tiles_nwaves = (ntiles_total + max_blocks - 1) / max_blocks;
const int tiles_efficiency_percent = 100 * ntiles_total / (max_blocks*tiles_nwaves); const int tiles_efficiency_percent = 100 * ntiles_total / (max_blocks*tiles_nwaves);