[SPARK-18223] Optimise PartitionedAppendOnlyMap implementation

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Duplicate
• Affects Version/s: 1.6.2, 2.0.1
• Component/s: Spark Core
• Assignee: Unassigned
• Reporter: Adam Roberts
• Created: 02/Nov/16 10:41
• Updated: 02/Nov/16 13:15
• Resolved: 02/Nov/16 13:15

Description

We can improve the PartitionedAppendOnlyMap implementation by taking advantage of of comparator wrapping, we have observed a 3% performance increase with both IBM's SDK for Java and OpenJDK on PageRank due to the inlining benefit

HiBench large used, pull request to follow

Issue Links

• duplicates: SPARK-18224 Optimise PartitionedPairBuffer implementation
• links to: [Github] Pull Request #15735 (a-roberts)