Ok, so my benchmarks weren't lying. While it's nice that the GCC devs are pushing optimizations (and from what I can tell 4.6 is pretty much all about better code performance) I'd personally prefer if they took some time to get their error reporting up there with llvm. I guess that performance is what the big customers want and this is why the (paid by big companies) devs are pushing optimizations so hard. We also have Intel hiring codesourcery to add specific optimizations to gcc for their core iX line.