Add a workaroud for issue in libcuda.so library #737
Merged
Conversation
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
|
I made change as requested so I am going to just go ahead and merge now. |
rolfv
pushed a commit
that referenced
this pull request
Jul 22, 2015
Add a workaroud for issue in libcuda.so library
|
does this need to go in to v2.x? |
|
@rolfv 👍 |
|
Yes. Will be submitting for 2.x. Will soak overnight. |
jsquyres
added a commit
to jsquyres/ompi
that referenced
this pull request
Aug 23, 2016
Fix the subarray and darray type creation.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.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.
We intermittently see that the check for GPU memory returns true when the memory is actually host memory. This workaround gets us past that until we get a fix in the CUDA library.
@bureddy Can you look at this? I inadvertently closed #728 so this takes its place. I made change as requested by @bureddy. It will not affect anything functionally but I can see how it seems to make sense to put the workaround before the CUDA initialization step.