Joseph K. Bradley created SPARK-11168:
-----------------------------------------
Summary: Explore precise behavior of treeAggregate
Key: SPARK-11168
URL: https://issues.apache.org/jira/browse/SPARK-11168
Project: Spark
Issue Type: Improvement
Components: MLlib, Spark Core
Reporter: Joseph K. Bradley
It would be worthwhile to make sure treeAggregate is efficient for small
numbers of partitions. We should examine its exact behavior in terms of the
structure of the tree aggregation on small problems sizes (maybe # partitions
from 1 to 20) and for different suggested depths.
A lot of use cases will be in this range, and if we can speed it up a bit, it
would be very worthwhile.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
