sunchao commented on PR #25491: URL: https://github.com/apache/datafusion/pull/25491#issuecomment-5835522083
Thanks @viirya. Rebased onto current main; the head is `8a964a88f4`. All checks that ran passed, with three extended-workflow jobs skipped. The full extended suite also passed locally. Yes, independent Comet tasks can build their broadcast hash tables concurrently across executor task slots. I updated the rationale to emphasize sharing the retained build/table and avoiding repeated materialization and hashing. This does not imply a proportional latency improvement; the cold microbenchmark does not establish end-to-end Comet latency savings. I also found that your earlier execution-time validation guard had been lost during simplification. It is restored, with a regression demonstrating that changing public join keys previously returned incorrect results and now rejects the incompatible prepared build. -- 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]
