mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-06 12:20:52 +01:00
f211818bc0
Add check to see if CUDA driver support multicast, as does in Symmetric Memory. Pull Request resolved: https://github.com/pytorch/pytorch/pull/147764 Approved by: https://github.com/syed-ahmed, https://github.com/yifuwang
11 lines
230 B
C++
11 lines
230 B
C++
#pragma once
|
|
|
|
// This file contains utility functions common for CUDA, which can be used by
|
|
// ProcessGroupNCCL or SymmetricMemory.
|
|
|
|
namespace c10d::cuda {
|
|
|
|
bool deviceSupportsMulticast(int device_idx);
|
|
|
|
} // namespace c10d::cuda
|