CUDA IPC maximum simultaneous peers limitation

[lilohuang]

Hi,

The simpleIPC.cu sample code describes the maximum simultaneous peers limitation for PCI-E cards. However, I couldn't find any detailed information from CUDA programming guide, is there an official documentation to describe the detail? Any CUDA API can be used to query the limitation?

https://github.com/NVIDIA/cuda-samples/blob/b312abaa07ffdc1ba6e3d44a9bc1a8e89149c20b/Samples/0_Introduction/simpleIPC/simpleIPC.cu#L38

Thanks, Lilo