You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Sean Owen (JIRA)" <ji...@apache.org> on 2017/02/27 10:18:45 UTC
[jira] [Resolved] (SPARK-19744) Find a better alternative to
netlib-java
[ https://issues.apache.org/jira/browse/SPARK-19744?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sean Owen resolved SPARK-19744.
-------------------------------
Resolution: Not A Problem
Hm, but why would it become easier if you forked the project, or, why would Spark do a better job reimplementing all this? the author, Sam, is really quite good at this. Sam's comment is that it's possibly hard to recompile for new or obscure architectures, but it's not work at all to support AMD64, x86, etc. Forking doesn't change the license issue, because it's the dependent libraries like libgfortran that are the problem. MLlib is not slow without this.
> Find a better alternative to netlib-java
> ----------------------------------------
>
> Key: SPARK-19744
> URL: https://issues.apache.org/jira/browse/SPARK-19744
> Project: Spark
> Issue Type: New Feature
> Components: Build, ML, MLlib
> Affects Versions: 2.1.0
> Environment: Fedora 25 Linux on x86-64
> Reporter: Demi Marie Obenour
>
> netlib-java is, by the author's own admission, [hard to build with native BLAS and LAPACK support|https://github.com/fommil/netlib-java/issues/83#issuecomment-114014185]. For licensing reasons, Spark cannot build with native BLAS and LAPACK support by default. This means that MLlib is very slow.
> A better solution would be to fork netlib-java (and make it easier to build) or to create an alternative that is easier to build. Ideally, the native component should be as easy to build as {{./configure && make && make install}}, and the Java component should be a standard Maven build.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org