Hi Andrew,
I suspect not and even if there are some around I would be very very
skeptical of the results. The Playstation 3 is a single precision chip and
while you can technically compile a single precision version of AMBER it is
unlikely to conserve energy and as such I would not trust the results
obtained from it. The same is true of any ports you see on other single
precision things like FPGA's etc. - be very very skeptical - unless it
passes the AMBER test suite it shouldn't be used.
I would also expect any benchmarks on PS3 to be pretty awful even if you
could get it compiled - the same is true of the Los Alamos Roadrunner
machine which 'essentially' is a bunch of double precision play stations
hooked together. Using the Cell processor is notoriously difficult because
it's underlying design is so poor. The ability to use the floating point
units is largely crippled by the internal chip design and hence no amber
developer, who are more focused on getting their research done, has bothered
to put in the effort to try to use such a chip. Hence on something like
Roadrunner you would just get the performance as if you were running on a
bunch of Opterons (the 1.8GHz host chips). Essentially the best performance
today for AMBER still comes from a decent cluster based system such as an
XT4 or Power5/6.
Sorry to disappoint you...
All the best
Ross
> -----Original Message-----
> From: owner-amber.scripps.edu [mailto:owner-amber.scripps.edu] On Behalf
> Of Andrew Ring
> Sent: Monday, December 22, 2008 4:41 PM
> To: amber.scripps.edu
> Subject: AMBER: Playstation 3 Benchmarks
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hello,
>
> Has anyone tried running Amber on a Playstation 3? If so, do you have
> any benchmark results? I am most interested in bench.factor_ix
> results, but any would do.
>
> Thank you,
> Andrew Ring
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
> iD8DBQFJUDOMlHj4gJAUE8oRAp+qAJkBW7r8M11zGj/2qgVmlota5IfpTQCcDI+f
> 71ckh4KSYVpViN5URVf6/08=
> =X2/e
> -----END PGP SIGNATURE-----
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
> to majordomo.scripps.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
to majordomo.scripps.edu
Received on Wed Dec 24 2008 - 01:14:51 PST
