rmuir commented on PR #12694: URL: https://github.com/apache/lucene/pull/12694#issuecomment-1770931189
> There is a test missing in TestVectorUtilSupport that compares the results of vectorized and standard impl. Also some basic tests using extreme vectors should be added due to overflows. > > As Robert says, I am quite sure that the current code overflows if vectorized if you have large values (like 0xFF). So please add a test that compares results (like we have for all other methods). Even if the code is fixed to always use `ZERO_EXTEND_B2I` and `ZERO_EXTEND_B2S` so that it does true unsigned math, there is the problem of the end result currently being defined as a 'signed integer'. It should be 'unsigned integer', but it leads to problems in java-land as we have to worry all code treats it correctly with `Integer.compareUnsigned` and that various "finalizer" functions acting on the result first cast to `long` and so on. This is not happening in the scalar versions right now either. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
