move all get_stream in sgl_kernel to c++ to reduce the launch overhead (#12521)

2025-11-02 13:15:05 -08:00
parent c9db79117f
commit 20315697f4
20 changed files with 61 additions and 94 deletions
--- a/sgl-kernel/python/sgl_kernel/utils.py
+++ b/sgl-kernel/python/sgl_kernel/utils.py
@@ -18,11 +18,6 @@ from typing import Dict, Tuple

 import torch

-
-def get_cuda_stream() -> int:
-    return torch.cuda.current_stream().cuda_stream
-
-
 _cache_buf: Dict[Tuple[str, torch.device], torch.Tensor] = {}