-
Notifications
You must be signed in to change notification settings - Fork 3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[GraphBolt][CUDA] Expose UniqueAndCompact
offsets.
#7789
[GraphBolt][CUDA] Expose UniqueAndCompact
offsets.
#7789
Conversation
To trigger regression tests:
|
@@ -299,10 +293,25 @@ UniqueAndCompactBatched( | |||
// Utilizes a sort based algorithm, the mapped id of a vertex part of the | |||
// src_ids but not part of the unique_dst_ids will be monotonically increasing | |||
// as the actual vertex id increases. Thus, it is deterministic. | |||
return UniqueAndCompactBatchedSortBased(src_ids, dst_ids, unique_dst_ids); | |||
auto results3 = |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why results3 & results4?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Because the world_size>1
code path is only supported for the Map based implementation. So the 4th argument is not returned by the sort based implementation. So we modify the output to fill the 4th output ourselves.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Since the code path works only when world_size=1
, we simply partition the whole unique_nodes
as [0, unique_nodes.size(0)]
to a single rank so that the output is valid for all codepaths, even though it won't be used.
Description
Towards implementing #7273.
Tests will be in a follow-up PR.
Checklist
Please feel free to remove inapplicable items for your PR.
Changes