I'll definitely implement a cache model to show cache behavior. I'm not sure branch prediction is realistic, though; I'm not actually parsing the code myself,

I'll definitely implement a cache model to show cache behavior. I'm not sure branch prediction is realistic, though; I'm not actually parsing the code myself, so finding the branch points would be a lot of extra work. I'm considering a scheme where you have to call an extra function inside each branching statement to get useful branch prediction results. Something along the lines of: for x in foreach([1...VA.length], "for1") y = 0 while branch(VA.gt(x, y), "while1") y++ if branch(y == x, "if1") break VA.insert(x, y) Where the `branch`, `foreach` functions are identities over their first argument, and they use the second argument to to track branching for a given branch point (i.e. you must uniquely name each branch, foreach line). A foreach would be a branch run arg0.length + 1 times, with arg0.length true branches followed by a single false branch. Sound at all reasonable?