1. 20

  2. 2

    An interesting topic – would have liked a more in-depth, generalized explanation though.

    Also, the author’s estimate of branch prediction accuracy is way off – 60% would be disastrous. Branch prediction accuracy for common workloads is typically well into the 90+ percent range – as a semi-arbitrary-but-perhaps-reasonably-illustrative datapoint, according to perf stat, my Haswell CPU’s branch prediction is 98.4% accurate for the build process of a ~10KLoC project I’m currently working on (most of that being GCC). The discussion of superscalar execution (and so forth) in the comments section isn’t terribly accurate either.