> Whoa! Nice write-up.
>
> Now, I don't have to be as careful about my conclusions as you do, so
> looking at your numbers... it looks like the performance you get on a
> dual-2Gig G5 is equivalent to what you'd see on a theoretical 5 Gig P4,
> if we extrapolate P4 floating performance linearly. Not too shabby.
>
> I was interested to see how close enabling the other processor came to
> doubling performance. The SMP architecture on these machines must be
> pretty good. Is there anything in particular that your software does
> to take advantage of multiple processors?
>
> -Patrick
>
Jet3D splits a problem up rather primitively -- if I were going to predict
noise at 20 observer locations (or microphone positions) on a dual-CPU
machine, the code would run two 10-observer jobs, assigning one per CPU (and
so on.... the same approach would apply to any number of CPUs up to 20).
In this case, the memory requirements are very low and the two CPUs probably
don't compete at all. In the future, I'd like to run much larger cases that
would test memory access, bus performance, etc. That's another area I would
expect the G5 to do much better than the G4.
In other cases I have looked at, with CFD codes that split problems up a
more typically using MPI, a dual G4 was about 1.8X better than a single G4,
so we definitely see a penalty there.
Craig
--
Dr. Craig Hunter
NASA Langley Research Center
AAAC/Configuration Aerodynamics Branch
email@hidden
(757) 864-3020
(Dual G4 - OS X)
_______________________________________________
scitech mailing list | email@hidden
Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/scitech
Do not post admin requests to the list. They will be ignored.
