Kernel sent to packet processor hangs indefinitely.

[Jorghi12]

Description:
A kernel is launched and sent to the packet processor to be processed. The kernel's completion signal handler never receives the 0 value (indicates completion), leading to an indefinite kernel hang.

HCC Trace:

Docker Image:
https://hub.docker.com/r/jorgdoku123/pytorch_hang_repro/

Python Reproduction:

import torch
src = torch.FloatTensor([1,2,3,4]).cuda(0)
dst = torch.LongTensor([[0]]).cuda(0)
torch.take(src,dst)