On Sun, 15 Sep 1996, Peter Berg wrote:

> 	AG> Yes, thanks. I worked for a company called D-Cubed, who
> 	AG> produce a variational-geometry constraint engine for CAD
> 	AG> systems. It currently runs to 240,000 lines of C++ and
> 	AG> breaks both the HP/UX and SunPro compilers on mild
> 	AG> optimisation. I'm going to compile it using gcc 2.7.2.1
> 	AG> on RiscBSD and see what happens :-)
> 
> If you get it to compile, dosn't you meen this will be terribly slow on the
> RPC ??
> I assume it will make heavily use of FP.
> Or will you do this only to test how strong gcc 2.7.2.1 is ?

It does make very heavy use of FP - I have a feeling it'll all but grind
to a halt under RiscBSD :-( It was more a test of the compiler than
anything else; I'm interested to see how gcc/ARM compares with the
heavyweights. 

If anything I was surprised at just how many serious commercial compilers
have broken optimisers. 

amit