[CK] Refactor GPU verification kernel to gather error stats on GPU #3551
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
johannes-graner
requested review from
ThomasNing,
afagaj,
andriy-ca,
aosewski,
asleepzzz,
bartekxk,
carlushuang,
cgmillette,
coderfeli,
geyyer,
illsilin,
poyenc,
qianfengz,
shumway,
tenpercent and
vidyasagar-amd
as code owners
Snektron
reviewed
Jan 12, 2026
Snektron
previously approved these changes
Jan 12, 2026
bartekxk
approved these changes
Jan 14, 2026
shumway
pushed a commit
that referenced
this pull request
Jan 15, 2026
…3551) * Refactor GPU verification kernel to gather erorr stats on GPU * Check if result is all zero * non-negative error count doesn't need custom Atomics * Remove unnecessary AtomicMaxFloat function * Simpler warp reduction, remove passed flag * Move verification header to include * Fix header path in test * Fix block reduction loop
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Outdated suggestions cannot be applied.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.
Proposed changes
Refactors the GPU verification to not exit early on errors, instead gathering all errors and returning a struct with error stats.
Checklist
Please put an
xinto the boxes that apply. You can also fill these out after creating the PR. If you're not sure, please don't hesitate to ask.clang-formaton all changed filesDiscussion
There is some slowdown for large tensors with very large number of errors (tested by setting reference tensor to 0 in
test_grouped_convnd_fwd). An initial implementation did not perform error statistic reduction in LDS, instead using atomic instructions for each error, this was extremely slow so the added complexity of LDS reduction is necessary.