upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-07-23 11:10:55 -05:00
Code Issues Packages Projects Releases Wiki Activity

cuda : disables launch_fattn PDL enrollment due to compiler bug (#23825)

Browse Source
This commit is contained in:
Andreas Kieslinger
2026-05-29 06:46:10 +02:00
committed by GitHub
parent 33c718db1f
commit 241cbd41d2
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

@@ -1153,8 +1153,8 @@ void launch_fattn(
GGML_ASSERT(block_dim.x % warp_size == 0);
const ggml_cuda_kernel_launch_params launch_params = ggml_cuda_kernel_launch_params(blocks_num, block_dim, nbytes_shared, main_stream);
ggml_cuda_kernel_launch(fattn_kernel, launch_params,
// disabled PDL enrollment for now due to a compiler bug.
fattn_kernel<<<blocks_num, block_dim, nbytes_shared, main_stream>>>(
(const char *) Q->data,
K_data,
V_data,