Add MULTI_REDUCE and HISTOGRAM Kernels #419
Open
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
MrBurmark
force-pushed
the
feature/burmark1/multireduce
branch
2 times, most recently
from
0cccafc
to
64a9461
Compare
MrBurmark
changed the title
MrBurmark
changed the title
MrBurmark
force-pushed
the
feature/burmark1/scan
branch
from
de0c81f
to
cc9bb82
Compare
This makes it easier to implement reproduction with reducers
Add the MULTI_REDUCE kernel which looks at performance for reducing into a runtime number of bins. In this case each iterate contributes to a single bin.
This is meant to be in some ways a simplification of the MULTI_REDUCE kernel and a way to compare with the performance of histogram implemenations in performance libraries like cub and rocprim
MrBurmark
force-pushed
the
feature/burmark1/multireduce
branch
from
64a9461
to
7ac439d
Compare
artv3
approved these changes
Mar 29, 2024
artv3
reviewed
Mar 29, 2024
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Outdated suggestions cannot be applied.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.
Add MULTI_REDUCE and HISTOGRAM Kernels
Add the MULTI_REDUCE and HISTOGRAM kernels which look at the performance for reducing into a runtime number of bins. In MULTI_REDUCE each iterate contributes to a single bin in the output like a histogram. HISTOGRAM allows us to have a comparison point with GPU performance libraries like cub and rocprim.
Which iterate maps to which bin is currently random, but this is not really a realistic use case. More realistic use cases are contiguous chunks of the domain mapping to the same bin, or even the entire domain mapping to a single bin.
The CPU parallel implementations are not optimized and simple use atomics into an array with no replication.