srowen edited a comment on issue #25545: [SPARK-28843][PYTHON] Set OMP_NUM_THREADS to executor cores for python URL: https://github.com/apache/spark/pull/25545#issuecomment-525966731 Maybe, though here still we'd actually overcommit cores in all cases, except `spark.executor.cores=1`, where it'd happen to use all cores. I don't think there is a scenario where this under-commits the machine's cores w.r.t. OpenMP. How about a mention in release notes? migration guide suggests some users would require a change to avoid a perf regression and I am just not sure that is even possible given this change.
---------------------------------------------------------------- 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] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
