Miniboxed Value Encoding
- Miniboxing (OOPSLA 2013) - the first miniboxing paper, describing the representation,
- Late Data Layout (OOPSLA 2014) - the foundation of miniboxing (also described in detail here) and
- Miniboxing the Linked List (tech report) - describes the miniboxing transformation for Scala collections.
The miniboxing plugin encodes all primitive values into long integers. To find out why, see the following resources:
Paper and Poster
We were notified by our peers that one of the times reported is impossible to reproduce. While this does not affect the paper's contributions, we would like explain the discrepancy in detail:
Table 3 lists the times taken by the List create microbenchmark for the generic case as:
- 16.7ms for the "Single context" case
- 1841ms for the "Multi context" case
The number from the "Multi context" case was very noisy (with a standard deviation of 1068ms as shown in the first version of the paper) and we are also not able to reproduce it. We think this number was affected by a job running on the server we used for benchmarking, which would explain the long running time and the huge standard deviation seen in the measurements. Our expectation is that the numbers for "Single context" and "Multi context" should be comparable. The measurements taken on the Graal virtual machine (in Table 4) exhibit this expected behavior. We apologize for not double-checking this number, misleading other authors who tried to reproduce it.
Talks
Devoxx UK 2015 (London, UK)
Miniboxing Warnings (PNWScala, Portalnd, OR)
Miniboxing Transformation (ScalaDays 2014 Berlin, Germany)
Other Resources
Comments
Comments are always welcome! But to make the best use of them, please consider this:
- If you have questions or feedback regarding the content of this page, please leave us a comment!
- If you have general questions about the miniboxing plugin, please ask on the mailing List.
- If you found a bug, please let us know on the github issue tracker.
Thanks! Looking forward to your messages!
comments powered by Disqus