В общем, оказалось, что все не так плохо в интерпретаторах, как многие думают. На Haswell'ах промах по branch prediction вообще в большинстве случаев незаметен в микроскоп.
По ссылке рассматриваются интерпретаторы JavaScript'а, Python'а и CLI
https://hal.inria.fr/hal-01100647/document
Branch Prediction and the Performance of Interpreters — Don’t Trust Folklore
... folklore has retained that a highly mispredicted indirect jump is one of the main reasons for the inefficiency of switch-based interpreters.
We show that the accuracy of branch prediction on in- terpreters has been dramatically improved over the three last Intel processor generations. This accuracy on Haswell, the most recent Intel processor generation, has reached a level where it can not be considered as an obstacle for performance anymore. We have also shown that this accuracy is on par with the one of the literature state-of-the-art ITTAGE (Indirect Target TAgged GEometric length predictor)
ЗЫ. Не знаю, может имеет смысл в «Философию» перенести.
