luhenry commented on pull request #30810: URL: https://github.com/apache/spark/pull/30810#issuecomment-749670059
Some more numbers. I ran a variant of https://github.com/apache/spark/blob/master/examples/src/main/scala/org/apache/spark/examples/ml/MultilayerPerceptronClassifierExample.scala to get some numbers comparing the f2j and vector implementations. Of 10 runs, the average training times are: - vector: 81s - f2j: 112.3s This is with the same dataset, same test, simply passing `--add-modules=jdk.incubator.vector` or not to the JVM. The JDK version used is build `16-ea+24`, and Spark has been compiled with `JAVA_HOME=/usr/lib/jvm/jdk-16+24 build/sbt package -Pvectorized -Djava.version=16`. You can find the output at https://gist.github.com/luhenry/5b54ca0bb3e64a99c2f8796826fd373f ---------------------------------------------------------------- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
