[jira] [Commented] (MATH-623) Use implementation of JAMA for ArrayRealVector operators

["Gilles (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MATH-623?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13068271#comment-13068271 ] 

Gilles commented on MATH-623:
-----------------------------

Hello Arne.

The _strongly_ recommended work method for modifying the code is to separate issues as much as possible. So, in this case, you propose an enhancement to the {{add}} method implementation in class {{ArrayRealVector}} but the diff contains many other modifications that might possibly be introduced later. If the changes are not interdependent, it is much easier to examine the merits of several changes one at a time.
Especially, the interface change of {{RealVector}} and {{SparseRealVector}} may be justified for their own sake, and do not seem to be directly dependent on the performance issue.

Thanks in advance for your patience...


> Use implementation of JAMA for ArrayRealVector operators
> --------------------------------------------------------
>
>                 Key: MATH-623
>                 URL: https://issues.apache.org/jira/browse/MATH-623
>             Project: Commons Math
>          Issue Type: Improvement
>    Affects Versions: 3.0
>            Reporter: Arne Plöse
>            Priority: Minor
>         Attachments: ArrayRealVector.diff
>
>
> For instance the add(double[] v) fist clones the array, and then adds all entries of this.data to the result.
> JAMA uses the following approach create a empty result[] the assign the sum of each entry to the result. this is approximately 10 -20 % faster.
> I will provide a patch with some more enhancements/cleanups for this.
> in the test look for XXX: 
> the first number is the jama time the second the current algorithm.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira