ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

CUDA: fix FA tg at long context for CC >= 8.9 (#13852)

Browse source
This commit is contained in:
Johannes Gäßler 2025-05-28 13:33:37 +02:00 • committed by GitHub
parent 26b79b6cb3
commit a68247439b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml/src/ggml-cuda/fattn-common.cuh
View file

@ -623,8 +623,8 @@ static __global__ void flash_attn_combine_results(
__builtin_assume(tid < D);
extern __shared__ float2 meta[];
if (tid < 2*parallel_blocks) {
((float *) meta)[threadIdx.x] = ((const float *)VKQ_meta) [blockIdx.z*(2*parallel_blocks) + tid];
for (int i = tid; i < 2*parallel_blocks; i += D) {
((float *) meta)[i] = ((const float *)VKQ_meta) [blockIdx.z*(2*parallel_blocks) + i];
}
__syncthreads();