FelixYBW commented on issue #7810: URL: https://github.com/apache/incubator-gluten/issues/7810#issuecomment-2578511921
> Do we plan to also do the opposite? - multiple task per thread. Currently, we depend on `spark.task.cpus` which is an integer, hence allows either one or more cpu cores per task. For I/O intensive tasks, it would be beneficial to fit more task per CPU. It's the over commit of CPU. We need to hack spark to support task.cpus=0.5 etc. Actually we can achieve the similar goal by adding multiple IO threads. you may try something like: --conf spark.gluten.sql.columnar.backend.velox.IOThreads=24 \ --conf spark.gluten.sql.columnar.backend.velox.prefetchRowGroups=1 \ --conf spark.gluten.sql.columnar.backend.velox.loadQuantum=268435456 \ --conf spark.gluten.sql.columnar.backend.velox.maxCoalescedBytes=67108864 \ --conf spark.gluten.sql.columnar.backend.velox.SplitPreloadPerDriver=0 \ -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
