ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-05 19:40:19 +00:00
Code Issues Packages Projects Releases Wiki Activity

Slightly better PP

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-09-01 09:17:19 +03:00
parent e2e5b270c5
commit bb21114ab4
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-cuda/fattn-mma-f16.cuh
View File

@@ -1235,7 +1235,9 @@ void launch_fattn_mma(
const int nblocks_stream_k = max_blocks;
const bool use_stream_k = cc >= CC_ADA_LOVELACE || tiles_efficiency_percent < 75;
//const bool use_stream_k = cc >= CC_ADA_LOVELACE || tiles_efficiency_percent < 75;
// On my RTX-4080 the above is slightly slower for PP. It would be useful to try and see what happens on Blackwell
const bool use_stream_k = tiles_efficiency_percent < 75;
blocks_num.x = use_stream_k ? nblocks_stream_k : ntiles_total;
blocks_num.y = 1;