Skip to content

[Fix] add block size logic for sm120 smem size #14311

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
+9 −1
Open

[Fix] add block size logic for sm120 smem size #14311

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
7004ee7
Update block size logic for sm120 architectures
koush Dec 2, 2025
386c7e8
fix up if nesting per review
koush Dec 2, 2025
9d3f94c
rename sm120 from consumer to workstation branding
koush Dec 2, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 9 additions & 1 deletion python/sglang/srt/layers/attention/triton_ops/extend_attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,7 +64,15 @@ def _get_block_sizes_for_extend_attention(Lq: int, Lv: int):
BLOCK_M, BLOCK_N = (64, 64)
num_warps = 4
else:
if _is_cuda and CUDA_CAPABILITY[0] >= 9:
if _is_cuda and CUDA_CAPABILITY[0] == 12:
# sm120 workstation Blackwell architecture (RTX Pro 6000) has a much smaller shared memory size (100K)
if Lq <= 128:
BLOCK_M, BLOCK_N = (64, 128)
elif Lq <= 256:
BLOCK_M, BLOCK_N = (64, 64)
else:
BLOCK_M, BLOCK_N = (32, 32)
elif _is_cuda and CUDA_CAPABILITY[0] >= 9:
# Hopper architecture (H100, etc.)
if Lq <= 256:
BLOCK_M, BLOCK_N = (128, 64)
Expand Down
Loading