Does current JIT optimize generated machine codes for branch prediction based on runtime statistics?
Some JVMs would compile Java byte code into native machine code. We know that there are lots of optimizations we could apply for that. Recently, I also learn that a branch operation may block the CPU ...
(I am asking this from a low latency perspective, but I guess the question applies in normal business domains). I am deciding whether to try and specialise in C++ or Java for low latency. My issue ...