On Wednesday 11 April 2001 20:43, Raab, Andreas wrote:
> Nah. I'd say for one thing it's poor x86 code to begin with. There are a
> few hacks that make a huge difference - for example, my (Win98) 700MHz P3
> gives me:
>
> '56,239,015 bytecodes/sec; 1,677,593 sends/sec'
>
> Note that the send speed scales with the machine speed but that the byte
> codes are actually faster on my (slower) machine. Somebody ought to look at
> that gnuify script and find out what's broken on Linux. There's a factor of
> two lurking in byte code speed right around the next corner.

Where? On my Celeron 466/533 Linux system, I get:

0 tinyBenchmarks '46343229 bytecodes/sec; 1553761 sends/sec'

then scaling by 700/533:
46343229 * 700.0 / 533         60,863,527
1553761 * 700.0 / 533            2,040,586

This seems to be in the range you're getting under Windows. Or even somewhat 
better, if you just scale by clock speed. I doubt the Celeron is any more 
efficient than the P3.

-- 
Ned Konz
currently: Stanwood, WA
email:     ned at bike-nomad.com
homepage:  http://bike-nomad.com