Baunsgaard opened a new pull request #1483: URL: https://github.com/apache/systemds/pull/1483
This commit change row aggregates in compressed space to first aggregate the dictionaries into vectors, and then populating the output. This makes it reduce the number of floating point operations needed. In the worst dataset Infini-MNIST it half execution time from 100 to 50 sec (uncompressed is at 8 sec) for 100 repetitions, and more importantly it reduce the number of instructions by 10x. Hopefully some better caching and threading will reduce it further in another commit. -- 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]
