Fix ORC reader when using device_read_async while the destination device buffers are not ready
#17074
Conversation
This file contains 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
Signed-off-by: Nghia Truong <nghiat@nvidia.com>
ttnghia
added
bug
ttnghia
changed the title
device_read_async when the destination device buffers are not readydevice_read_async while the destination device buffers are not ready
vuule
reviewed
Oct 14, 2024
| // Instead, it may use some other stream(s) to sync the H->D memcpy. | ||
| // As such, we need to make sure the device buffers in `lvl_stripe_data` are ready first. | ||
| if (!stream_synchronized) { | ||
| _stream.synchronize(); |
There was a problem hiding this comment.
should be just sync before the loop?
There was a problem hiding this comment.
Yes that works, but would that be inefficient if we don't have any device read? We will sync after the loop if here is only host read.
vuule
approved these changes
Oct 14, 2024
davidwendt
approved these changes
Oct 14, 2024
|
/merge |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
This fixes a bug in ORC reader when
device_read_asyncis called while the destination device buffers are not ready to write in. In detail, this bug is becausedevice_read_asyncdoes not use the user-provided stream but its own generated stream for data copying. As such, the copying ops could happen before the destination device buffers are being allocated, causing data corruption.This bug only shows up in certain conditions, and also hard to reproduce. It occurs when copying buffers with small sizes (below
gds_threshold) and most likely to show up with settingrmm_mode=async.