"Carl Gundel" <carlg at libertybasic.com> is claimed by the authorities to have written:

> I ran the following two benchmarks:
> 
> Time millisecondsToRun: [
>  1 to: 10000 do: [ :i | i printString ] ]
> 
> This gave me slightly more than 12000 milliseconds.  :-(
> 
> Time millisecondsToRun: [
>  1 to: 1000000 do: [ :i |  ] ]
> 
> This gave me slightly more than 900 milliseconds.
On my seven year old 206MHz SA desktop machine (ie a 32 bit wide damp
string memory bus, well, calling it a bus is an exageration, more a
handcart) I get 2200mS and 700mS. The three major differences I know of
between your machine and mine are:-
a) I have a 'real' StrongARM 110, you have a StrongARM1110 with much
smaller caches (and they're already tiny on a 110)
b) you have (almost certainly) a much faster memory architecture
c) your code is compiled via a M$ complier rather than the ancient Acorn
one that I use.
I guess it just goes to show how utterly meaningless to real world
performance the simple cpu MHz and memory bandwidth numbers are!

With the latest images and the latest VMs I just find morphic bearable
on my Acorn; lots more work is needed but I can use it without screaming
in frustration. I've never yet run an XScale based machine (though I may
just buy an Iyonix) but the evidence seems to be that the compiler needs
to understand the implications of the longer pipeline in order to not
disappoint. I'd rather have a Hallah cpu if possible, but they ain't
shipping yet :-(

tim

-- 
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
Klingon Code Warrior:- 1) "Behold, the keyboard of Kalis!  The greatest Klingon code warrior that ever lived!"