Paul Barton-Davis wrote:
> 
> >It's a combination of things: for one, the P-II isn't very speedy
> >converting between integer and floating types.  Other architectures
> >don't take nearly the hit from the conversion.  Your benchmark does
> 
> just one clarification: i didn't write this benchmark. i don't know
> who did.

Me neither, but I remember seeing it on a page about rtcmix. Said page
seems to have since disappeared. (Info about rtcmix is now at:
http://www.music.columbia.edu/cmix/)
 
I stole the basic idea -- adding up oscillators, incrementing a count
onscreen which you watch until the output glitches -- for my csound
benchmark orcs (which are on the benchmark page).

Argh! I just tried my benchmark orcs on my new system, and I notice that
linrand and trirand no longer work as they used to. Around version 3.49
I could use linrand to put a random pitch into an i-rate variable; but
now (unofficial linux version 3.53) it seems that linrand gets the same
seed for every note, so I get the same random pitch every damn time. I'm
submitting a bug-report to the linux csound folks...