Optimized build for SPARC Ultra IV processor (MP) machine

Hi,

we are going to run ROOT on a SUN E2900 (8 Ultra IV processors) to create root files
of approx. 1Gb size. Since there is a limitation on the time spend for the creation of these
files, I did some investigations how one can get a optimized build for such a machine (more precise, for the Ultra IV processor 1.2GHz).
After numerous investigation the situtaion looks like follows:

Sice we are just creating trees (and branches, etc.) I took the bill.C example from:
root.cern.ch/root/html/examples/bill.C.html to do some benchmarking (to not be
dependend from X11).

All test have been done using ROOTversion 4.02/00 which showed the following results:
Binary from root.cern.ch: 350 rootmarks
Compiled by myself (CC 5.3 with -fast -xtarget=ultra -xarch=v8plus) 390 rootmarks
Compiled by myself (CC 5.6 with -fast -xtarget=ultra4 -xarch=v8plus) 400 rootmarks
Running on an AMD Opteron (~2.4GHz) I got something like 1200 rootmarks!!

So to come to the point: We need to get at least a factor of 2 to make the writting not
to hard to programm (multithreaded writting into one file?!)!
So my question is if one knows how I can get this factor of two somewhere or …?!