Well, I got myself a copy of IBM's xlc compiler for the G5 (not the beta
as it expired) and decided to do some more (seriously unofficial)
benchmarking on my dual G5 2ghz.
The long and the short of it is that the official Mac version as
downloaded from the Povray website still wins at 32 minutes, 5 seconds.
I downloaded the 3.6.1 source tarball and compiled it using xlc two ways
-O3 -qarch=g5 -qaltivec
-O5 -qarch=ppcv -qaltivec -qnostrict -qfloat=nonans -qignerrno
The surprise is that the less optimized version (-O3) ran it in 35
minutes, 43 seconds. The more optimized version (-O5) did it in 39
minutes, 47 seconds.
I also used gcc with -mpowerpc-gpopt -mpowerpc64 -faltivec and got 38
minutes, 37 seconds. Compare this with the way it's compiled when
installed via Fink and you get 39 minutes, 35 seconds.
Post a reply to this message