[
https://issues.apache.org/jira/browse/FLINK-3181?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15062462#comment-15062462
]
ASF GitHub Bot commented on FLINK-3181:
---------------------------------------
GitHub user vasia opened a pull request:
https://github.com/apache/flink/pull/1467
[FLINK-3181] [gelly] avoid unnecessary messages in SSSP examples and
library method
I added a simple check in the messaging function to avoid sending
unnecessary messages during the first superstep of vertex-centric SSSP.
Initially, the workset contains all vertices, which means that all of them will
execute the messaging function during the first iteration. However, only the
source needs to generate messages in the first superstep; all other vertices
would produce unnecessary messages with infinite-value distances.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/vasia/flink FLINK-3181
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/1467.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #1467
----
commit 07525f4e0934b6421085999ac58fb5901e325554
Author: vasia <[email protected]>
Date: 2015-12-17T15:11:58Z
[FLINK-3181] [gelly] avoid unnecessary messages in SSSP examples and
library method
----
> The vertex-centric SSSP example and library method send unnecessary messages
> during the first superstep
> -------------------------------------------------------------------------------------------------------
>
> Key: FLINK-3181
> URL: https://issues.apache.org/jira/browse/FLINK-3181
> Project: Flink
> Issue Type: Improvement
> Components: Gelly
> Affects Versions: 0.10.0
> Reporter: Vasia Kalavri
> Assignee: Vasia Kalavri
>
> The result is correct, but the current implementation has unnecessary message
> overhead in the first superstep. The Messaging function should only produce
> messages from the source during the first superstep. This can be fixed with a
> simple check of the vertex value.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
