I'm glad you like the idea! I noticed that F2J is written in C/C++, and
unfortunately I'm not terribly familiar with C\C++ and haven't written
parser generators before either. Java may have the ability to run on GPUs
in the not too far future (there's a project called Sumatra that is
underway), so my hope is that by leveraging Lambdas LAPACK would be able to
run on Java using GPU hardware when the time comes.
I also noticed that the PLASMA project has boasted that they have more
efficient algorihtms -- should I look at it as a starting point instead, or
would LAPACK be a better starting point?
Thanks so much,
On Thu, Jan 30, 2014 at 12:27 PM, julie langou <julie@Domain.Removed> wrote:
This is a great idea.
This would require some work, first to update the F2J to integrate the new
Java functionalities, and update the translate tool to integrate the
changes to LAPACK.
At the moment, we do not have the resources to work on such a project and
it is out of our current scope of work. (Mr Seymour is no longer working at
Nonetheless, LAPACK as you know, is mostly contribution based. So if you
know somebody that is willing to work on this project, the LAPACK team will
do our best to support his effort.
On Jan 30, 2014, at 9:26 AM, Jim Bethancourt <jimbethancourt@Domain.Removed>
A coworker of mine is using LAPACK and is using JNI calls to call the
Fortran version. I noticed that the JLAPACK project is availalbe, but it
looks kind of old. I was wondering if it would be of any interest to
regenerate a modern version of JLAPACK based on the November 2013 release
of LAPACK, decompile it, and update it to use the new forEachParallel,
Lambda, and Stream features that are being added in Java 8. It would be
interesting to see the before and after performance increases and API call
simplifications as well, and it could also be a great way to test & explore
the new Java 8 features.
Lapack mailing list
-------------- next part --------------
An HTML attachment was scrubbed...