[Pdl-porters] SIMD in PDL

David Mertens dcmertens.perl at gmail.com
Mon Sep 26 04:31:19 HST 2011


Kartik -

PDL supports trivial parallelization, i.e. parallelization in which
parallel computations do not need to talk to one another. In other
words, PDL can easily parallelize matrix-marix addition, but not
matrix-matrix multiplication. It looks like PLASMA would provide a
great framework that actually parallelizes individual linear algebra
calculations that PDL is not presently able to do well. As such, I
think that Perl or PDL bindings for PLASMA would be a worthy effort
and would satisfy a niche that PDL does not yet handle well.
Obviously, I would vote for PDL bindings, but I would understand if
you wanted to take the route of standard Perl-level bindings. In that
case, integration to play nicely with the non-PDL GSL bindings might
be worth considering, though I know nothing about those bindings.

David

On Sat, Sep 24, 2011 at 9:52 AM, Kartik Thakore
<thakore.kartik at gmail.com> wrote:
> Hey Guys,
>
> Is there support for SIMD in PDL? SIMD is
>
> http://en.wikipedia.org/wiki/SIMD
>
> If not is there an interface I can make for a new XS module I am
> considering on making on the PLASMA library?
> http://icl.cs.utk.edu/plasma/software/index.html
>
> Regards
>
> --
> Kartik Thakore <thakore.kartik at gmail.com>
>
>
>
> _______________________________________________
> PDL-porters mailing list
> PDL-porters at jach.hawaii.edu
> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>



-- 
Sent via my carrier pigeon.




More information about the PDL-porters mailing list