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

Support TMA multicast on B with m_grouped_gemm_contiguous.

Browse Source
This commit is contained in:
yukuai
2025-04-18 15:38:04 +08:00
parent 4499c4ccbb
commit 2752a67aad
4 changed files with 71 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
deep_gemm/jit_kernels/gemm.py
View File

@@ -146,10 +146,9 @@ def get_best_configs(m: int, n: int, k: int, num_groups: int, num_sms: int,
best_tma_multicast_config = (1, True)
# Try to multicast on the larger block side first
is_dense_gemm = (not is_grouped_contiguous) and (not is_grouped_masked)
is_multicast_legal = {
'A': is_tma_multicast_legal(n, best_block_n, 2, num_sms),
'B': is_tma_multicast_legal(m, best_block_m, 2, num_sms) and is_dense_gemm,
'B': is_tma_multicast_legal(m, best_block_m, 2, num_sms) and (not is_grouped_masked),
}
for i in ('A', 'B') if best_block_m > best_block_n else ('B', 'A'):
if m >= 512 and is_multicast_legal[i]: