mboehm7 commented on PR #2376: URL: https://github.com/apache/systemds/pull/2376#issuecomment-3698735914
LGTM as well. Thanks for the patch and experiments @Biranavan-Parameswaran, thanks for the review @janniklinde, and sorry for the delay. During the merge I just added a few TODOs (for ourself in the future) to experiment with multiple tasks per thread which usually improves load balance, and to call a common kernel from both single- and multi-threaded operations (i.e., single-threaded simply calls the multi-threaded kernel with start=0 and end=numRows). The special case for clen=1 is safe because our in-memory blocks only support numRows/numCols of max integer, which means the allocated dense block will always be contiguous. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
