[MPFR] GMP+MPFR to GPU?

[Rick C. Hodgin]

Paul,

How well suited to parallel / vector processing are GMP's and MPFR's
algorithms?  If, for example, we setup N identical operations for
parallel computing (adds, multiplies, cosines, whatever), how would the
existing design handle the vectorization?

- Rick

On Wed, 2010-12-01 at 09:00 +0100, Paul Zimmermann wrote:
> Rick,
> 
> > From: "Rick C. Hodgin" <foxmuldrster at yahoo.com>
> > Date: Wed, 01 Dec 2010 01:12:36 -0600
> > 
> > I am realizing as I begin using GMP (and MPFR) how much I need to have
> > this library ported to GPU architecture for additional processing speed.
> > 
> > I wanted to ask if there are any efforts underway toward this end?  If
> > not, would somebody be interested in striking up a dialog with me to
> > port GMP (and MPFR) to ATI and NVIDIA GPUs?
> > 
> > I am willing to do all programming under the current license, as an
> > off-shoot of GMP (and MPFR).  But it's doubtful it would succeed if I
> > did it on my own.
> > 
> > I'm asking for help in this endeavor.  Thank you.
> > 
> > - Rick C. Hodgin
> 
> MPFR's efficiency very much relies on the efficiency of GMP's mpn functions,
> thus the first thing would be to port mpn to GPU.
> 
> Porting mpn to GPU is surely possible, but an efficient port is more tricky.
> 
> Paul Zimmermann