I simply want to share this empirical study on the LLVM -O3 pipeline on the PolyBench suite. I don't want to bore you with too many details that are already in the paper. Any feedback is welcome :D