On Sun, 5 Jan 1997, Peter Berg wrote:
> > > 3) Does povray run correctly yet ? I recall that it used to coredump. 
> 
> No ! The problem is that gcc is still not able to compile it correct. I used 
> gdb to see what could be changed, but it looked to complex for me :(
> The compiled binary crashes on calculating line 24 of blob.pov.
> The problem is in the sub "intersection elimination"

 I don't think this is a gcc problem. The compiled code runs fine on a 
machine with a FPA.  IIRC the problem is with the kernel not trapping FP 
exceptions correctly. You can't stop it from producing a FP exception on 
a divide by zero etc... and crashing the program.  Something like this is 
what is happening to povray.

   Mark mentioned some time ago that to fix this would result in reduced 
FP performance, and I don't know what the status on it has been since then.

> 
> > > 4) Does someone fancy doing some timing tests on the FP performance under
> > > StrongARM ? The question on FP performance really ought to have a MFLOPS
> > > figure to quantify it.

   Watch out when measuring MFLOPS on RiscBSD as it counts time spent in 
the FPE as system time, whereas MFLOPS expects it to count as user time.

   I modified MFLOPS once to make it use total (i.e. system+user) time, 
and it runs about 10 times quicker than a 30MHz Arm6 (Not too surprisong 
there eh?) but still depressingly slow compared to most other things.

  Sorry, but I don't have the source or results on me riught now, but 
it's easy to see how to modify it yourself.

> > > 7) Is there any 16-bit sound support yet - either a simple beep or full
> > > audio ?

  I rewrote mkbeep  a few weeks back to add 16 bit support. 
Hopefully Mark'll release it soon.

Ale.