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

Merge pull request #44 from sazczmh/main

Browse Source 
Performance: Configuration algorithms tuned to minimize the impact of tail effects, now up to 1402 TFLOPS
This commit is contained in:
Chenggang Zhao 2025-03-10 13:08:03 +08:00 committed by GitHub
commit ba1e93a5c7
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 11 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
deep_gemm/jit_kernels/gemm.py
View File

@ -101,7 +101,14 @@ def get_best_configs(m: int, n: int, k: int, num_groups: int, num_sms: int,
if m >= 1024 and is_tma_multicast_legal(n, best_block_n, 2, num_sms) and num_groups == 1:
best_num_tma_multicast = 2
return best_block_m, best_block_n, best_num_stages, best_num_tma_multicast, best_smem_size
# Recompute the minimal number of SMs required
# NOTES: less L2 cache usage and less GPU frequency drop
num_waves = get_num_waves(best_block_m, best_block_n)
num_min_sms = ceil_div(ceil_div(m, best_block_m) * ceil_div(n, best_block_n) * num_groups, num_waves)
num_min_sms = ceil_div(max(num_min_sms, num_sms - 8), best_num_tma_multicast) * best_num_tma_multicast
assert num_min_sms <= num_sms and is_tma_multicast_legal(n, best_block_n, best_num_tma_multicast, num_min_sms)
return num_min_sms, best_block_m, best_block_n, best_num_stages, best_num_tma_multicast, best_smem_size
def gemm_fp8_fp8_bf16_nt(lhs: Tuple[torch.Tensor, torch.Tensor],
@ -151,7 +158,7 @@ def gemm_fp8_fp8_bf16_nt(lhs: Tuple[torch.Tensor, torch.Tensor],
# Auto-tuning with compilation
global includes, template
num_sms = get_num_sms()
block_m, block_n, num_stages, num_tma_multicast, smem_size = get_best_configs(m, n, k, 1, num_sms)
num_sms, block_m, block_n, num_stages, num_tma_multicast, smem_size = get_best_configs(m, n, k, 1, num_sms)
args = (lhs, lhs_scales, rhs, rhs_scales, out, m, torch.cuda.current_stream(), num_sms, smem_size)
runtime = jit_tuner.compile_and_tune(
name='gemm_fp8_fp8_bf16_nt',

4
deep_gemm/jit_kernels/m_grouped_gemm.py
View File

@ -84,7 +84,7 @@ def m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(lhs: Tuple[torch.Tensor, torch.Ten
# Auto-tuning with compilation
global includes, template
num_sms = get_num_sms()
block_m, block_n, num_stages, num_tma_multicast, smem_size = get_best_configs(m, n, k, 1, num_sms,
num_sms, block_m, block_n, num_stages, num_tma_multicast, smem_size = get_best_configs(m, n, k, 1, num_sms,
is_grouped_contiguous=True)
args = (lhs, lhs_scales, rhs, rhs_scales, out,
m_indices, m, num_groups,
@ -158,7 +158,7 @@ def m_grouped_gemm_fp8_fp8_bf16_nt_masked(lhs: Tuple[torch.Tensor, torch.Tensor]
# Auto-tuning with compilation
global includes, template
num_sms = get_num_sms()
block_m, block_n, num_stages, num_tma_multicast, smem_size = get_best_configs(expected_m, n, k, num_groups, num_sms)
num_sms, block_m, block_n, num_stages, num_tma_multicast, smem_size = get_best_configs(expected_m, n, k, num_groups, num_sms)
# Extra checks for TMA store
if num_groups > 1 and m > block_m: