My guess would be that branch misprediction does have an impact on interpreted language, but much less. If bytecode instructions take on average 20 CPU cycles to execute, and the branch misprediction penalty is 50 CPU cycles, the relative cost of a misprediction is much smaller than in compiled code.
Comments
My guess would be that branch misprediction does have an impact on interpreted language, but much less. If bytecode instructions take on average 20 CPU cycles to execute, and the branch misprediction penalty is 50 CPU cycles, the relative cost of a misprediction is much smaller than in compiled code.
the counterpoint is that a lot of those 20 instructions are also branches