Github user dlwh commented on the pull request:
https://github.com/apache/incubator-spark/pull/575#issuecomment-35105330
Just to follow up on Breeze performance: in the latest snapshot, we are
consistently faster than JBlas and Mahout in @mengxr's benchmarks.
On Fri, Feb 14, 2014 at 1:54 AM, MLnick <notificati...@github.com> wrote:
> I for one would prefer to use netlib-java, whether it is via MTJ or
Breeze.
> I've used both and find the MTJ API pretty good and with good sparse
> support (though this was a few years ago admittedly). Breeze has a nicer
> more "Matlab" DSL but as I mentioned there's no reason not to wrap MTJ in
a
> lighter-weight DSL that is not heavy on implicits, if for performance
> reasons it is desirable.
>
>
> On Fri, Feb 14, 2014 at 11:37 AM, Sam Halliday <notificati...@github.com
> >wrote:
>
> > @MLnick <https://github.com/MLnick> I have it on good authority (from
> the
>
> > author of JBLAS) that he consider netlib-java to exceed his original
> > goals for JBLAS. I am utterly confused why commons-math would accept
> JBLAS
> > but reject netlib-java as a backend, given that JBLAS doesn't even have
a
> > java backend and they were concerned that the netlib-java java backend
> > could not be compiled without classfile manipulation.
> >
> > I'll hold fire on the license request until you decide based on
technical
> > merit. I don't anticipate any problems: in fact I believe he would be
> > delighted that MTJ is continuing.
> >
> > --
> > Reply to this email directly or view it on GitHub<
> https://github.com/apache/incubator-spark/pull/575#issuecomment-35069073>
>
> > .
> >
>
> â
> Reply to this email directly or view it on
GitHub<https://github.com/apache/incubator-spark/pull/575#issuecomment-35070192>
> .
>
