LLVM 3.3 To Introduce SLP Vectorizer

With additional vectorization work to be done by LLVM, the compile times of software will increase.

The SLP vectorizer didn't cause any performance changes for C-Ray but attempting to use the "-fslp-vectorize-aggressive" compiler switch led to a noticeable regression in performance.

The Smallpt performance also regressed with the aggressive SLP option.

Overall, there isn't too much to get excited about for LLVM/Clang compiler users right now with the SLP Vectorizer. While there's hope that this Superworld-level parallelism will provide speed-ups for straight-line code, right now the benchmark results of LLVM/Clang from an Intel Core i7 system running Ubuntu Linux x86_64 yielded little performance changes. Of more benefit right now is the loop vectorizer that is now enabled by default with "-O3" for the forthcoming LLVM 3.3 release.