You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@systemml.apache.org by "Matthias Boehm (JIRA)" <ji...@apache.org> on 2018/03/31 01:42:00 UTC
[jira] [Created] (SYSTEMML-2218) Avoid unnecessary repartitioning
after RDD parallelize
Matthias Boehm created SYSTEMML-2218:
----------------------------------------
Summary: Avoid unnecessary repartitioning after RDD parallelize
Key: SYSTEMML-2218
URL: https://issues.apache.org/jira/browse/SYSTEMML-2218
Project: SystemML
Issue Type: Task
Reporter: Matthias Boehm
There are cases, where spark mapmm instructions to a repartition of the input RDD in order to ensure output partition sizes <2GB. If the input originates from a RDD parallelize this causes unnecessary shuffle because the parallelize could already bring it into the right number of partitions.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)