Improve performance of CUDA kernels generated from manual_sparse_jacobian #4093
Conversation
This file contains hidden or 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
daverumph
force-pushed
the
dr/gpu_perf/manual_sparse_jacobian_1
branch
from
59848ad to
5b7e47e
Compare
imreddyTeja
reviewed
Nov 10, 2025
Co-authored-by: Teja Reddy <[email protected]>
Co-authored-by: Teja Reddy <[email protected]>
Co-authored-by: Teja Reddy <[email protected]>
daverumph
commented
Nov 11, 2025
daverumph
commented
Nov 11, 2025
imreddyTeja
approved these changes
Nov 11, 2025
Member
|
The regression test breaks - is the behavior change expected? |
Member
|
Also could you squash the commits? Thanks! |
Member
|
The results have changed because |
Hoping to address small floating point differences
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Outdated suggestions cannot be applied.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.
Purpose
Use cached values to avoid doing extra work in the kernel.
Content
The changes selectively add back use of cached variables that were removed en masse by Charlie a while back. Upon profiling with NVidia's NSight tools, we now understand that some of that caching is valuable to avoid "global" (from the GPU's perspective) memory reads.