Thomas Tauber-Marshall has posted comments on this change. Change subject: IMPALA-2932: Extend DistributedPlanner to account for hash table build cost ......................................................................
Patch Set 3: (1 comment) http://gerrit.cloudera.org:8080/#/c/4098/2/fe/src/main/java/com/cloudera/impala/planner/DistributedPlanner.java File fe/src/main/java/com/cloudera/impala/planner/DistributedPlanner.java: Line 461: lhsTree.getCardinality() * ExchangeNode.getAvgSerializedRowSize(lhsTree)); > you're not guaranteed to have computed rhsDataSize at this point Sorry about that. It seems like this might be a bug with the existing code, though? Since it means that we don't do the hash table size check at line 489 in situations where we have a rhs cardinality estimate but no lhs num nodes, since we don't calculate the broadcast cost even though we could calculate the size of the rhs in this situation. I submitted a new version with both of these things fixed, but of course I can do it a different way if it was intentional that we don't do the hash table size check for the broadcast in situations where we can't calculate the broadcast cost. -- To view, visit http://gerrit.cloudera.org:8080/4098 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I03a0f56f69c8deae68d48dfdb9dc95b71aec11f1 Gerrit-PatchSet: 3 Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-Owner: Thomas Tauber-Marshall <[email protected]> Gerrit-Reviewer: Marcel Kornacker <[email protected]> Gerrit-Reviewer: Matthew Jacobs <[email protected]> Gerrit-Reviewer: Mostafa Mokhtar <[email protected]> Gerrit-Reviewer: Thomas Tauber-Marshall <[email protected]> Gerrit-HasComments: Yes
