Giters

NVIDIA / warp

A Python framework for high performance GPU simulation and graphics

Home Page:https://nvidia.github.io/warp/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

FeatureRequest: Multistream

jc211 opened this issue · comments

commented

I think the idea of a SyncedMultiStream would be a nice addition to warp. My understanding is that this enables quick fork and join strategies. An implementation from another NVIDIA repo is found here:
https://github.com/NVlabs/tiny-cuda-nn/blob/28ca991f99b44d10387d73077c07ccfdd7f96275/include/tiny-cuda-nn/multi_stream.h

commented

Thanks @jc211, I hadn't seen this before - indeed looks like it could be a useful pattern. It seems to rely only on standard CUDA stream behaviors, so I think it should be possible to implement in terms of the current Warp stream and event APIs. @nvlukasz for viz.