[Beowulf] Re: Q&A: BeOS and IBM Cell processor

Mark Hahn hahn at physics.mcmaster.ca
Mon Jan 23 09:39:40 PST 2006


> q1) How about the BeOS? ... any reports on attempts or successes in  
> using BeOS on clusters and multiprocessor systems?
> 
> Considering that the latest iterations of BeOS seems to have been  
> specifically targeted at multiprocessor, multitasking projects, this  
> might seem to be a natural ... and not just a toy of garage shops

yeah, but why do you think the OS matters much?  in a cluster,
you just want it to get out of the way and let the program run.
does BeOS do something dramatically better?


> q2) Anyone examining the IBM Cell processor? ... Is this processor  
> worthy of consideration for clusters and multiprocessor systems?
> 
> Same, same for the IBM Cell processor = targeted at the future  
> multiprocessor markets and not just game systems (like Sony  
> Playstation 3).

Cell is not impressive: it's a PPC chip with some attached processors
for doing FP-intensive, low-data crunching (like game rendering.)
I tried hard to find any sign of interesting new computing paradigms,
and completely failed.  it's pretty straightforward: you run a normal
OS on the (very normal) PPC core, and do DMA-like transfers to get 
stuff in and out of the attached processors.  no real ISA cleverness,
no Big Vision (like, say the i432, hah!)

yes, if you put a bunch of FPUs on a chip you get lots of flops.  wow.
just as with GPUs, the reported flop rates are far more impressive 
if you're using single-precision.  I don't know of many non-video
codes that use 32b floats...




More information about the Beowulf mailing list