tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-30 06:03:37 -04:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: use tensor cores for MMQ (#7676)

Browse Source 
* CUDA: int8 tensor cores for MMQ (legacy quants)

* fix out-of-bounds writes

* __builtin_assume -> GGML_CUDA_ASSUME

* fix writeback returning too early
This commit is contained in:
Johannes Gäßler
2024-06-10 11:45:13 +02:00
committed by GitHub
parent af4ae502dd
commit 1f0dabda8d
7 changed files with 550 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml-cuda/fattn-tile-f16.cu
View File

@@ -43,7 +43,7 @@ static __global__ void flash_attn_tile_ext_f16(
const int ne1,
const int ne2,
const int ne3) {
#if FP16_AVAILABLE
#ifdef FP16_AVAILABLE
//In this kernel Q, K, V are matrices while i, j, k are matrix indices.
const int ic0 = (blockIdx.x / parallel_blocks) * ncols; // Index of the Q/QKV column to work on.