For now we only mark innermost loops for the loop vectorizer. We could later also mark not-innermost loops to enable the introduction of openmp parallelism. llvm-svn: 202854
