DragonFly BSD
DragonFly kernel List (threaded) for 2004-12
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: IBM "Cell" processor


From: Dave Leimbach <leimySPAM2k@xxxxxxx>
Date: Thu, 16 Dec 2004 07:20:35 -0800
Cancel-lock: sha1:8aTatQB6y8pi3sizeG5nK84DO+o=

esmith <esmith@xxxxxxxxxxxx> writes:

> Garance A Drosihn wrote:
>> February".  I don't think computers will start shipping (not in
>> any volume) by then.  PS3's, perhaps, but not general-purpose
>> computers.
>
> After these remarkable claims of perfomance it's probably best to take
> a "wait and see" approach but, the fact that workstations do exist
> right now ( the URL you posted speculated that developers already have
> cell workstations ) makes this a curiosity, because some of the
> hardware features do seem to match well with some of the DF goals.

I think this "cell" technology is somewhat related to PCA project:

http://www.morphware.org.  Being able to dynamically reconfigure the "cpu"
for different applications.  I think this means you need to think of the 
processing unit in a new light and that cell is directly related to this.

Also the performance numbers you probably see from it are most likely
highly dependent on locality of reference which means "streaming" applications.

There is an awful lot of research in this area right now...  I used to get
funding to work on this stuff a couple years back but it wasn't high enough
for the company I work for to continue working on the project.  There are 
companies working on compilers for all sorts of stuff like "cell" technology
to graphics card GPUs.  All within the realm of high performance computing.

-- Dave



[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]