Re: Problems with GCC on 386

[Stanislaw Stepien <starmaker@semestr-1.t4.ds.pwr.wroc.pl> writes:]

> hi there
>I have one question: why floting point operations are needed to compile
>programs ? Which part of compiler uses floting points operations ?

From memory, there's this one register usage calculation which uses floating
point instead of integer/fixed point. In the very earliest version of gcc I
ever ported to VSTa, I got a patch out of the old djgpp (gcc for DOS) port
which changed it to not use float. Later versions of djgpp stopped patching
gcc, so eventually VSTa supported the FPU and I stopped patching, too. This
is mostly OK, but occasionally bites some poor soul using a 80386 or
80486sx.

If somebody wanted, they might be able to dredge up old VSTa (or djgpp) gcc
source, extract the technique, and see what it would take to bring it
forward. I'd be happy to fold a successful patch back in and make that the
gcc for the distribution. Compilers using the FPU for basic algorithms
seems wrong to me, too.

Andy Valencia
Received on Sat May 22 06:30:31 1999