coutinho at dcc.ufmg.br
Sun Aug 24 19:02:37 PDT 2008
Today GPUs have two levels of parallelism. They are a set of SMID processors
that use SPMD parallelism between them.
In NVidia GPUs, for example, the stream multiprocessor has only on
instruction decoder and all eight stream processors execute instructions
issued by decoder in SMID. The parallelism between stream multiprocessors is
SPMD, sharing data through global memory.
As common GPU workloads are very regular, this saves the area of seven
In ATI GPUs it's almost the same, but the stream multiprocessor is called
SMID core and has 16 stream processors.
2008/8/23 Mikhail Kuzminsky <kus at free.net>
> BTW, why GPGPUs are considered as vector systems ?
> Taking into account that GPGPUs contain many (equal) execution units,
> I think it might be not SIMD, but SPMD model. Or it depends from the
> software tools used (CUDA etc) ?
> Mikhail Kuzminsky
> Computer Assistance to Chemical Research Center
> Zelinsky Institute of Organic Chemistry
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Beowulf