Lpc2000

I'm starting design work on a robotics controller that requires
lots of floating point cos(), sin(), and sqrt().

I use 16 bit Freescale MC9S12 processors in the current version of 
the design.  With the fastest possible clock and the Codewarrior 
compiler it handles floating point multiplies in a decent 8 
microseconds, but slips to ~400 microseconds for cos(), and ~250 
microseconds for  sqrt().  This puts me at the minimum timing specs 
for the design.

Would appreciate hearing any numbers you might have for cos(), sqrt), 
etc. on the smaller ARMs, and the compiler, processor, and clock 
speed that produced them.

Thanks!

Bill Tondreau
http://www.kupercontrols.com

Mileage will vary depending oon compiler, memory map etc.

I use gcc and I did some benchmarks a while ago (over a year back) 
using an old version of gcc patched with a FP update. On a 30MHz ARM7 
I got 2 secs for 10,000 sqrt() calls.

The best thing to do, IMHO, is to write a benchmark and run it on a 
the micro of your choice.



--- In lpc2000@yahoogroups.com, "tonalbuilder2002" 
<twentiethwave@h...> wrote:
> 
> I'm starting design work on a robotics controller that requires
> lots of floating point cos(), sin(), and sqrt().
> 
> I use 16 bit Freescale MC9S12 processors in the current version of 
> the design.  With the fastest possible clock and the Codewarrior 
> compiler it handles floating point multiplies in a decent 8 
> microseconds, but slips to ~400 microseconds for cos(), and ~250 
> microseconds for  sqrt().  This puts me at the minimum timing specs 
> for the design.
> 
> Would appreciate hearing any numbers you might have for cos(), 
sqrt), 

> etc. on the smaller ARMs, and the compiler, processor, and clock 
> speed that produced them.
> 
> Thanks!
> 
> Bill Tondreau
> http://www.kupercontrols.com