Re: Reverse engineering of Intel branch predictors

To me, that looks like the bytecode is insufficiently expressive, if
an often-repeating sequence occurs :-)

The "often-repeating sequence" is usually because there is a loop in the
bytecode program.

Are you suggesting that bytecode languages should shun loops, presumably
by replacing such a loop by a dedicated bytecode instruction?
That might be an option for those rare bytecode interpreters that are
sophisticated enough to generate program-specific bytecode instructions,
but those are sliding toward JIT compiler-land.

For your run-of-the-mill bytecode interpreter which has ~200
instructions defined manually once and for all, "often-repeating
sequence" are the lay of the land, I think.


        Stefan