Github user hbdeshmukh commented on the issue:
https://github.com/apache/incubator-quickstep/pull/224
@zuyu There is a threshold to decide whether we perform the group by
aggregation in a partitioned way or not. If the estimated number of groups
(provided by the optimizer) is greater than the threshold, we go for
partitioned aggregation. In TPC-H Q10 for SF100 the estimate comes out to be
50k and the default threshold is also 50k. Therefore we were not able to
perform parallel partitioned aggregation. When I tried to force the aggregation
to be partitioned, the addressed bug was exposed. With the changes in this PR,
now the aggregation is partitioned and hence the performance improvement.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
