Investigate CoalescedHashPartitioning
https://github.com/apache/spark/commit/b53c170679f added a new CoalescedHashPartitioning case class. We need to investigate whether the RAPIDS Accelerator needs to be updated to handle this new partitioning.
See https://github.com/apache/spark/commit/81639090622 for changes that were needed to the CPU BroadcastHashJoinExec that are probably relevant to the changes likely needed for the GPU version.
related to #9753
See apache/spark@81639090622 for changes that were needed to the CPU BroadcastHashJoinExec that are probably relevant to the changes likely needed for the GPU version.
This commit updated the outputPartitioning logic in BroadcastHashJoinExec, but we do not currently implement this method in the GPU equivalent class. I filed https://github.com/NVIDIA/spark-rapids/issues/10609 to discuss whether we should also implement this, but as @revans2 pointed out, AQE will call the CPU version of this method anyway during replanning and before we replace these operators, so maybe we don't really need them.