Skip to content

[perf] Avoid dtype promotion sync in mamba_get_block_table_tensor #34870

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

Merged
vllm-bot merged 3 commits into from
Feb 20, 2026
Merged

[perf] Avoid dtype promotion sync in mamba_get_block_table_tensor #34870

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
dbe11bd
linear_attn
hl475 Feb 19, 2026
8be9c08
Merge branch 'main' into linear_attn
hl475 Feb 20, 2026
60375ba
Merge branch 'main' into linear_attn
DarkLight1337 Feb 20, 2026
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
8 changes: 6 additions & 2 deletions vllm/v1/attention/backends/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -855,8 +855,12 @@ def mamba_get_block_table_tensor(
(seq_lens - 1) // kv_cache_spec.block_size,
min=0,
)
# Use int32 for arithmetic to avoid dtype promotion overhead,
# then convert to int64 for gather (which requires Long indices)
offsets = torch.arange(
1 + kv_cache_spec.num_speculative_blocks, device=block_table.device
1 + kv_cache_spec.num_speculative_blocks,
device=block_table.device,
dtype=torch.int32,
)
indices_to_gather = start_indices.unsqueeze(1) + offsets
indices_to_gather = (start_indices.unsqueeze(1) + offsets).to(torch.int64)
return torch.gather(block_table, 1, indices_to_gather)