[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Occam and the parallel Playstation
What got me thinking the Cell might be a natural for POP is the Blachford
That's version 2, fixed after arstechnica's Hannibal attacked it on some
details, but it still has the main points that attracted my attention:
supposedly it requires "concrete programming" and has a very local
resource structure. This should be extremely friendly to a CSP/occam
architecture, and not to much else.
Damian, I'd be really interested in being kept in the loop on occam for
the Cell. (I did try getting the head of research at Sony US interested in
a similar project, and he was interested, but it did not pan out.) I think
you're exactly on the right track to use interrupts (and, I presume, DMA)
to implement the primitives. I did something similar long ago with the DOS
PC. A "default harness" that exploits the fastest characteristics of the
Cell should be possible, if Blachford is at all on the right track.
> Yes, it's true, we are indeed working on a version of the transterpreter
> for the cell. We're also looking at generating C and using the
> transterpreter core as the runtime for occam programs. While much of the
> work is still in the early stages, we do have bits working. There
> are two papers which are going to be part of this years CPA describing
> what we have done to date, hopefully by the time we get to CPA development
> will have progressed.
> I'll drop a note to the list when we have something that can be released,
> at the moment its still under quite heavy development and isn't too
> I'm really happy to see that people are interested in it though, helps
> motivate me :).
> As to the cell's communications system, I don't know about it being
> designed with CSP but it can be programmed in a 'csp-friendly' way.
> The hardware doesn't enforce any csp-like behaviour, if any of it seems
> csp-like then I think thats a coincidence. CSP-ness needs to be done
> in software, with a bit of help from interrupts that the hardware can
> generate as well as some harware enforcable blocking/non-overwriting
> communications registers.
> On Thu, 3 Aug 2006, Chalmers, Kevin wrote:
>> The Cell Processor's communication system was also supposedly designed
>> using CSP. I do remember reading that the Transterpreter people
>> (http://www.transterpreter.org <http://www.transterpreter.org/> ) wanted
>> to use the Cell Processor. If they had contact or not I'm not sure.