Re: [Discuss-gnuradio] ps3 performance

[Robert McGwier]

This and other questions.  You are not doing something right.  Just 
using the spe's for fft's we will see a huge speed up in gnuradio.  I 
can run filter, NCO, baud clock and carrier recovery, and equalizer on 
the cell on 40 Mbaud signals.  Trying getting that out of a (Intel) cola 
nut.

We can't support that  data rate into the ps3 but I can into the cell 
servers at work.  The PS3 will run like crap if you are running a 
windowing environment, even one of the lightweight ones because of small 
memory. That said, we are going to be able to really increase our signal 
processing throughput by selective use of the SPE's and decent planning 
our use.

Bob


Eric Blossom wrote:
> On Tue, Oct 30, 2007 at 09:44:39AM +0200, Juha Vierinen wrote:
> > BTW, terrasoft sells ps3 pre-installed with linux (and a root password
> > supplied). I have never even connected a monitor to my ps3.
> >
> > But as to performance, has anyone measured decent performance on a
> > ps3? I am starting to think that it is not possible to achieve
> > anything near the theoretical peak performance. I am often getting
> > better performance with a intel core2 than on the 6+1 processors that
> > come with Cell. Intel c compiler often comes up with a faster program
> > just by autovectorizing a standard c program --  and this is compared
> > to something hand vectorized and paralellized for Cell. It could be
> > just that I'm not at all good at optimizing code, but this is also
> > true for some of the examples that come with the IBM SDK (eg. the
> > euler solver). The only thing that I have found to be impressive was
> > the cell optimized fft.
>
> (1) Are you ordering and aligning your SPE instructions such that you're
>     getting dual issue?
>
> (2) Are you providing branch hints?  A mispredicted branch stalls for
>     11 cycles.
>
> Eric
>
>
> _______________________________________________
> Discuss-gnuradio mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/discuss-gnuradio


--
AMSAT Director and VP Engineering. Member: ARRL, AMSAT-DL,
TAPR, Packrats, NJQRP, QRP ARCI, QCWA, FRC. ARRL SDR WG Chair
“An optimist may see a light where there is none, but why
must the pessimist always run to blow it out?” Descartes