dongjoon-hyun commented on PR #60195: URL: https://github.com/apache/spark/pull/60195#issuecomment-6703581394
Thank you, @viirya. You're right. I updated the PR description to say that the cost is small: a broadcast is deserialized once per executor, and the freed 1 MiB page goes back to the `HeapMemoryAllocator` pool. I also removed the paragraph about the 64 MiB page in the released versions. There, when the data fit in one page, `read()` allocated a page of the same rounded size, so it reused the freed array right away. -- 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]
