Fix kernel_dot_part2 for smaller BLOCKSIZE #87
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
This kernel is called with both BLOCKSIZE 1024 and 256. For the 256 case, the first two if-statements are trivially true, but then we access entries in LDS outside of the allocated LDS space. We leverage the constexpr-if idiom in C++17 to determine at compile time if we need the additional checks at larger BLOCKSIZE.
So hopefullly we don't make the same mistake again.
ntrost57
approved these changes
May 30, 2025
768C
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Outdated suggestions cannot be applied.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.
Builds on #86 and is the primary motivation from my end for moving to c++17, namely to be able to use
if constexpr(). This is not the most general solution, but does fix the existing use-cases in the dot product code; I did not scour the other parts of the code for similar instances. Commit message lays out the issue in detail.If we don't think this is enough impetus (together with rocPRIM deprecating c++14) to move to c++17 then I can cook up a different solution to this problem that only uses c++14.
Tested the fix on the offending setup, as well as existing (functioning) setups.