@chandlerc
It reminds me how many programming languages got a little bit faster when processors started using branch target prediction for dynamic jumps/calls, which weren't optimized earlier because they are rare in C code.