I am looking for a Linux compiler for C/C++ code that supports AMDs new "Bulldozer" architecture and produces efficient binaries for the Interlagos series Opterons. This seems to be a bit difficult because of the peculiarities of the Bulldozer microarchitecture. While AMD has a whitepaper with some details, I would like to see some independent analyses. The relevant paper from HeCToR focuses mostly on job placement and scheduling, which is an area we already investigate. So, who can recommend a good compiler comparison for Bulldozers running Linux? Does anyone have well-described benchmarks?