tkonolige commented on pull request #6671: URL: https://github.com/apache/incubator-tvm/pull/6671#issuecomment-708649133
The problem with using fork is that it is unsafe on macOS or when using the CUDA api. The pytorch developers recommend not using fork with CUDA: https://pytorch.org/docs/master/notes/multiprocessing.html#cuda-in-multiprocessing. It looks like we have to choose something that works over something that is fast. I also don't think it really matters that spawning is slower than forking. We are not spawning that many processes. And if the overhead from spawning is big enough to cause a significant slowdown, then we shouldn't have been using processes/threads in the first place. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
