Re: [oc] video processors

["Marko Mlinar" <Marko.Mlinar@campus.fri.uni-lj.si>]

> > I know my above statement was a bit provocative :) but I believe that
VLIWs
> > and DSPs are more complicated than or2k.
>
> I agree. I prefer simpler designs like MISC and MOVE (TTA) processors
> as well. For those not familiar with these terms:
>
>    http://www.ultratechnology.com/mup21.html
>
>    http://cardit.et.tudelft.nl/MOVE/
huh... you are giving so much URLs... I can't read all that in so little
time :)

> > I mean - or2k is enough powerful
> > and enough general purposed to decompress videos, sound, etc and at the
> > same time much simpler. We calculated, that 3x3 or2k matrix with 8-16k
cache
> > would be on 4mm2 :) And frequency >1GHz would be easably reached.
> > I think that these things count too.
>
> These numbers wouldn't be easy to achieve with 0.8 micron technology,
> which is what I will be using after I debug my design with FPGAs. With
> something more modern then you are right - the or2k would be a great
> core to have. But you said work on it had been stopped, right?
Work on or2k core was stopped (or just delayed) because we wanted to make
faster RISC GP proccesor (let's say for PCs). We had opinion that it is
would
bring too much trouble for market - new programing techiques, compiler,...
We estimated GP market is not ready for that now (at least we don't have
enough power to change anything). And also applications that needs heavy
calculation power has better special purpose processors like supercomputers
or parallel comp. But I said that this design would be most suitable for
mathematically complex and general aplication with high ILP.
I think audio and video filters would be perfect application for or2k.
I hope I made things clear.
Beside its speed I think another thing is important for OpenIPCores:
its simplicity - nobody wouldn't build heavy duty DSP uP with this kind
funding and this type of development...

BTW: I am interested in your design. Could you pass more detailed info?

regards,
    Marko