LLVM Clang 3.8 Compiler Performance Benchmarks

First up with the HMMer benchmark we see a nice improvement in performance using this latest Clang compiler code.

Clang 3.8 appears to have a regression with the Monte Carlo test of SciMark2.

The other SciMark2 tests indicated no change in performance.

Himeno sadly seems to regress as well with this upcoming compiler update.

With LLVM/Clang continuing to add more optimizations, the compile times are slowing down slightly but still should be much faster than the current compiler speed of GCC.