[Haskell] The Cell processor and Haskell

Mads Lindstrøm mads_lindstroem at yahoo.dk
Thu Mar 16 16:22:06 EST 2006


Hi all

It is often claimed that Haskell's referential transparency would make
it an excellent choice for a high-level parallel language. Furthermore,
IBM, Sony, and Toshiba's Cell processor is in great need of a high-level
parallel language. Thus, they seem to be a perfect match.

See http://www.research.ibm.com/cell/ and
http://en.wikipedia.org/wiki/Cell_processor for more about the Cell
processor.

My knowledge of parallel programming is limited, but I would still like
to know, what the wiser amongst of us, thinks of making a Haskell
compiler for the Cell processor. It is not that I consider implementing
one. It is just that I am curious. 

Is the Cell processor better suited, as a target for a parallel Haskell
compiler than Intel, AMD, and Sun's multicore approaches? What I am
asking here is, if it is easier to make use of the Cell's closely
coupled but limited SPEs or it is easier to make use of the loosely
couple but full-featured parallelism of multicore approaches.

Is it realistic to think that a Haskell compiler targeting the Cell
processor, could make use of it's parallelism without requiring huge
changes to the source code of ones already existing meant-for-GHC
programs?

/Mads Lindstrøm




More information about the Haskell mailing list