I compared compilation time with clang version 3.5 that ships with latest ubuntu and the one I built (with --enable-optimized --enable-shared). The results are:
3.5 from ubuntu: 2,60s user 0,14s system 99% cpu 2,766 total
3.6 (a33591f612fa56e7c83fe348e7ca6c32778aacb0): 3,92s user 0,14s system 99% cpu 4,078 total
Did run the same compilation few times with the same compiler to avoid some cold caches.
This means almost 50%. Any reason for this? Any other build hints for more aggressive optimizations?