On Tue, 26 Oct 1999, Herbert Wengatz 42850 wrote:
>
> Hi Alan,
>
> +> > More importantly though (because we already know that Intel isn't
> +> > exactly error-free on the FPU front), will codes that are P-II/P-III
> +> > optimised for floating-point run quickly on the Athlon?
> +>
> +> Double precision yes. For single precision low accuracy work it'll do
> +> just nicely but rewriting it to use 3dnow! will make it scream.
> +>
> +> Stream_d for a 500MHz Athlon. Code not optimised for Athlon.
> +>
> +>
> +> Function Rate (MB/s) RMS time Min time Max time
> +> Copy: 384.2000 0.0424 0.0416 0.0431
> +> Scale: 377.8219 0.0467 0.0423 0.0590
> +> Add: 452.5828 0.0555 0.0530 0.0641
> +> Triad: 431.6395 0.0598 0.0556 0.0726
>
> Sorry, the same results for an 500 MHz PIII would be nice here, too.
> Just for a quick comparison. :)
Here you go:
Function Rate (MB/s) RMS time Min time Max time
Copy: 306.9751 0.1056 0.1042 0.1125
Scale: 187.2528 0.1714 0.1709 0.1733
Add: 235.2814 0.2044 0.2040 0.2052
Triad: 240.4930 0.2000 0.1996 0.2016
>From a dual 500 MHz P-III, Intel N440BX motherboard.
Don Holmgren
Fermilab
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/dmentre/smp-howto/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
