DeepSeek/DeepGEMM
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepGEMM synced 2025-06-26 23:15:49 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

add notes2

Browse Source
This commit is contained in:
yukuai 2025-04-22 15:56:40 +08:00
parent 8a92750027
commit 80f1cfc630
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
deep_gemm/include/deep_gemm/scheduler.cuh
View File

@ -48,7 +48,7 @@ struct Scheduler {
// NOTES: For the case where the total number is an odd number of blocks, the last block requires special barrier processing.
// Here, we need each cluster to have exactly two blocks.
DG_STATIC_ASSERT(kNumTMAMulticast <= 2, "Scheduler does not support > 2 TMA multicast");
if (num_blocks_in_group == 1 and n_block_idx == kNumNBlocks - 1 and m_block_idx == num_aligned_m_blocks - 1
if (kNumTMAMulticast == 2 and num_blocks_in_group == 1 and n_block_idx == kNumNBlocks - 1 and m_block_idx == num_aligned_m_blocks - 1
and num_aligned_m_blocks % 2 == 1 and kNumNBlocks % 2 == 1)
return false;
return true;