I compiled recent releases for OS X and used gcc's optimizations for G4 processors. This gives about 10 % performance when I try the benchmark files provided on Mark Stock's pages. I also submitted the result to him. The opt-line in the makeall script here is as follows:
set opt="-O2 -DSPEED=200 -mcpu=7400 -mtune=7400 -malign-natural -funroll-loops -fstrict-aliasing -fsched-interblock -falign-loops=16 -falign-jumps=16 -falign-functions=16 -falign-jumps-max-skip=15 -falign-loops-max-skip=15 -ffast-math -mpowerpc-gpopt -fstrict-aliasing"
Those who want to try my binaries can get them directly from the following address:
3.6 patch 1 release:
HEAD release from 18th january:
The files will be linked on the binary-pages I maintain on lgrobe.port5.com (those are linked from radiance-online.org), but the host is currently down and I have to wait that my provider brings it back
CU, and have fun, Lars.