Stochastic Program Optimization
By encoding constraints of transformation correctness as terms in a cost function, and using a Markov Chain Monte Carlo sampler to explore the space of all possible code sequences, we are able to generate aggressively optimized versions of a given target code sequence.
