frozenleaves/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-10-20 14:53:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

[ROCm][Kernel] Using the correct warp_size value

Browse Source
This commit is contained in:
Gregory Shtrasberg
2025-02-05 22:15:08 -05:00
committed by GitHub
parent 75404d041b
commit 5b19b93082
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
csrc/moe/moe_align_sum_kernels.cu
View File

@ -207,8 +207,8 @@ __global__ void sgl_moe_align_block_size_kernel(
__shared__ int32_t shared_counts[32][8];
__shared__ int32_t local_offsets[256];
const int warp_id = threadIdx.x / WARP_SIZE;
const int lane_id = threadIdx.x % WARP_SIZE;
const int warp_id = threadIdx.x / 32;
const int lane_id = threadIdx.x % 32;
const int experts_per_warp = 8;
const int my_expert_start = warp_id * experts_per_warp;