Baunsgaard opened a new pull request #1017: URL: https://github.com/apache/systemds/pull/1017
This PR contains code fro compressed Matrix Mult. Before the parallelization strategy for compressed matrix mult was a row at a time, this left optimization potential on the table. This PR optimize the techniques to show a 2x performance gain on matrix mult with the covtype dataset. the number after mml refer to the number of rows in the other matrix left multiplied with the compressed covtype dataset. ... DATA , RUN , TYPE , TIME ms , REP covtype , MatrixVector mml25 , ula , 92.488 , 250 covtype , MatrixVector mml25 , cla , 40.308 , 250 covtype , MatrixVector mml50 , ula , 143.116 , 250 covtype , MatrixVector mml50 , cla , 64.196 , 250 covtype , MatrixVector mml100 , ula , 230.508 , 250 covtype , MatrixVector mml100 , cla , 132.800 , 250 covtype , MatrixVector mml200 , ula , 243.976 , 250 covtype , MatrixVector mml200 , cla , 132.900 , 250 covtype , MatrixVector mml400 , ula , 257.564 , 250 covtype , MatrixVector mml400 , cla , 134.736 , 250 ... ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
