Change spatial traversal to fix perf regression from #897 #976
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
dalg24
approved these changes
Dec 7, 2023
|
I've run the bvh benchmark on Saturn for Serial and CUDA, and surprisingly, it showed no difference in times. Maybe that's why we did not catch the regression. |
|
What is "Saturn" again? |
It's a bit of mystery for me. Possibly. I'm pretty sure I played around with different traversals in #897 but could not find a good one. It could be that Power9 was regressing or something. |
One of the University of Oregon OACISS center systems. |
Open
|
HIP build failure was "No space left on device" |
|
All pass (HIP did not run). |
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Outdated suggestions cannot be applied.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.
Fix perf regression from #897
On Saturn with the standard HACC 37M problem (eps=0.042), the traversal time goes down (compared to master (fbd2584)):
minpts = 2 (
query+clustertimer):FDBSCAN: 0.147 -> 0.114
FDBSCAN-DenseBox: 0.154 -> 0.126
minpts = 10
FDBSCAN
neigh: 0.058 -> 0.057query: 0.137 -> 0.115FDBSCAN-DenseBox
neigh: 0.057 -> 0.057query: 0.161 -> 0.129This patch should restore most of the lost performance due to APIv2 refactoring:
