Observation for <<The partial reverse if-conversion Framework for balancing...>>
This paper deals with some back-track optimization. We can see from this
paper that the ordering of applying these optimization is crucial to get good result.
It's a good idea to first let the compiler do agressive optimization and later patch on it.
It's surpring to me that for many benchmarks the Hyperblock framework actually leads to
negative speedup.

Zhigang Hu