Improve performance of knn traversal #357
Merged
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
#neighbors = 10#neighbors = 1So, it seems that for Serial it is quite important to check the nodes that one gets from stack for the distance, and not do extra work. For Cuda, though, it seems preferential to not do that. |
Host runs a lot faster with (node, distance) stack, while Cuda without. Checking the distance really matters for small values of k, like nearest neighbor.
|
Summit results (master e8bdf3a vs branch dd0bf48): #neighbors = 1#neighbors = 10#neighbors = 100 |
|
OK, I'm reasonably happy with this PR now. Cuda is much faster, even for k = 1. Serial is about the same. The only downside is a slight increase for OpenMP for low k. An interesting observation: POWER9 showed no difference, but my Intel workstation significantly improved. We should add non-POWER9 performance testing to our toolbox (Intel or AMD)performance. Maybe ALCF? |
aprokop
changed the title
dalg24
reviewed
Aug 13, 2020
No. You think that 4-byte alignment could be worse than 8-byte? |
|
I meant alignas(16) PairNodePtrDistance stack[64]; |
I see. I thought you meant adding padding to In any case, there is a lot of tinkering that could be done in this PR. However, as any tinkering would require rerunning many performance tests, I would like to keep it to minimum, and only for strictly necessary things. |
dalg24
approved these changes
Aug 14, 2020
| heap.popPush(leaf_pair); | ||
| if ((int)heap.size() == k) | ||
| radius = heap.top().second; | ||
| } |
There was a problem hiding this comment.
The duplication of code is unfortunate.
There was a problem hiding this comment.
Yeah, I was really annoyed by the duplication. The only way to clean it up would be to to move part of the logic inside heap itself, but it's another can of worms.
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.
No description provided.