Hello, I am learning how to use nvshmem from your code. But I have a small problem and hope to get your help.
When multiple pes get into a dispatch/combine function, I noticed that you used wait_until for synchronization, but if a pe hangs at this time, wait_until will hang forever and cannot reach the subsequent device/stream_synchronize.
I don't know how pplx solves this problem? Thanks in advance for your responses!