Parallel Haskell: 2-year project to push real world use

Tyson Whitehead twhitehead at gmail.com
Sat May 8 12:21:14 EDT 2010


On May 3, 2010 08:04:14 Johan Tibell wrote:
> On Mon, May 3, 2010 at 11:12 AM, Simon Peyton-Jones
> <simonpj at microsoft.com>wrote:
> > In truth, nested data parallelism has taken longer than we'd hoped to be
> > ready for abuse :-).   We have not lost enthusiasm though -- Manual,
> > Roman, Gabi, Ben, and I talk on the phone each week about it.  I think
> > we'll have something usable by the end of the summer.
>
> That's very encouraging! I think people (me included) have gotten the
> impression that the project ran into problems so challenging that it
> stalled. Perhaps a small status update once in a while would give people a
> better idea of what's going on. :)

Most will likely see this on Planet Haskell, but it seem worthwhile mentioning 
as there seems to be quite a bit of interest in the technology.  Manuel just 
finished posting quite a nice talk given by Simon PJ recently in Boston

http://pls.posterous.com/simon-peyton-jones-on-data-parallel-haskell

Related to some of the questions asked at the talk, I would be curious to hear 
any comments regarding adding support for processor level SIMD vectorization 
(e.g., the SSE{1,2,3} instructions on the x86) in conjunction with NDPH.  It 
would seem conceptually as simple as having "ideal" vector primitive 
operations and coding the basic operations (e.g., sumS) in terms of those.

This would then presumably have a very positive impact on the the existing 
Vector, Repa, etc, libraries as well, and would seem quite a bit easier than 
trying to recognize opportunities for these instructions via loop 
analysis/whatever and insert them after the fact?  Or is the recognition 
option felt to be "better" (i.e., in the sense it would also be applicable in 
other situations as well) and easily done with the big data-flow hammer?

I see there is a track ticket regarding SIMD instructions

http://hackage.haskell.org/trac/ghc/ticket/3557

so I'm guessing at least part/all of the answer is just the time to do the 
grunt work to add in the support to CMM and the native code generators.

Cheers!  -Tyson
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part.
Url : http://www.haskell.org/pipermail/glasgow-haskell-users/attachments/20100508/65caf8f8/attachment.bin


More information about the Glasgow-haskell-users mailing list