Date   : Tue, 14 Feb 2006 14:18:55 +0000 (GMT)
From   : Sprow <info@...>
Subject: Re: MOS 3.20 v 3.50

Hi,
Seeing as my ARM7 coprocessor is plugged in on test, I might as well run
your tests too.

Personally I take most benchmarks with a pinch of salt since you almost
always end up with a really contrived piece of test code. For example
running the count-to-ten-thousand test from
  http://www.beebmaster.co.uk/BeebHelp/6502Test.html
I couldn't beat the 35s figure since basically that's how long it takes for
the VDU drivers in the OS to scroll and print the numbers out - the parasite
processor has to sit waiting for the Tube FIFO to empty.

In article <003201c6302b$641a4cd0$3201a8c0@...>, David Hunt wrote:
> I did some performance tests using a bit of BBC Basic;
>
> Simple trig. function into floating point variable
> --------------------------------------------------
>
> 65C102@...        65C102@...      
> 68.01s    33.17s
>
> Z80A@...  32016@...
> 38.21s    24.89s

ARM7 coprocessor@... (cache on/cache off)
0.12s / 3.13s

> Insert value into array
> -----------------------
>
> 65C102@...        65C102@...
> 4.86s             2.36s
>
> Z80A@...  32016@...
> 6.86s             4.12s

ARM7 coprocessor@... (cache on/cache off)
0.02s / 0.8s

Feel the speed,
Sprow.