commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Phil Steitz (JIRA)" <>
Subject [jira] Closed: (MATH-269) Use F2J translated BLAS/LAPACK
Date Wed, 14 Apr 2010 00:37:56 GMT


Phil Steitz closed MATH-269.

> Use F2J translated BLAS/LAPACK
> ------------------------------
>                 Key: MATH-269
>                 URL:
>             Project: Commons Math
>          Issue Type: New Feature
>    Affects Versions: 2.1
>            Reporter: Sam Halliday
> This is part of the proposal to integrate MTJ, netlib-java and commons-math which has
been well received on the developers' list, @see,-matrix-toolkits-java-and-consolidation-tt23537813.html
> Commons Math currently uses hand-crafted BLAS/LAPACK-like methods, whereas the F2J has
maintained a full Java translation of the BLAS and LAPACK fortran code, and is applicable
to most of the scientific code built on top of those libraries (e.g. ARPACK). It enables maintenance
to BLAS/LAPACK to be handled by the netlib team, even for the Java translation. BLAS and LAPACK
are industry standard APIs.
> This proposal is to incorporate the translations into the commons-math project. A follow-up
RFE will to define the API to be used in place of the existing code (i.e. incorporate netlib-java).
I do not recommend making the F2J translations a part of the public API.
> The code translated by F2J follows the same BSD licence as the netlib libraries. I understand
this is compatible with the Apache licence.
> F2J translations can be directed to any package name, currently they output in the "org.netlib.{blas,lapack}"
domain, but commons-math may prefer to use "org.apache.commons.math.netlib".
> There is a caveat: the translation of Fortran code introduces a language feature not
available in Java - the GOTO statement. Java translations are post-compile processed by a
"javab" tool (distributed with F2J) which augments the bytecode and allows the code to behave
as the Fortran original. This rules out the possibility of simply including the source code
in the commons-math tree, unless the javab tool becomes part of the build process.
> The alternative is to maintain a custom binary (built using F2J from netlib sources),
or to use a maven dependency.

This message is automatically generated by JIRA.
If you think it was sent incorrectly contact one of the administrators:
For more information on JIRA, see:


View raw message