Lazily initialize CUDA devices (take 2)#613
Merged
soumith merged 1 commit intotorch:masterfrom Nov 26, 2016
Merged
Conversation
Previously, cutorch would initialize every CUDA device and enable P2P access between all pairs. This slows down start-up, especially with 8 devices. Now, THCudaInit does not initialize any devices and P2P access is enabled lazily. Setting the random number generator seed also does not initialize the device until random numbers are actually used.
Member
|
thanks! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Previously, cutorch would initialize every CUDA device and enable P2P
access between all pairs. This slows down start-up, especially with 8
devices. Now, THCudaInit does not initialize any devices and P2P access
is enabled lazily. Setting the random number generator seed also does
not initialize the device until random numbers are actually used.
I've updated the Storage copy code to delegate the Tensor copy code. This
fixes the issues with p2p not being enabled and adds proper inter-GPU
synchronization (see #612)