Does any JIT/AOT/hot code optimization/techniques/compilers/runtime takes into account whether the branch prediction is saturated and try to recompile to go branchless
In general branchless is better for branches that can't be predicted 99.something %, saturating the branch prediction like this benchmark isn't a concern. The big concern is mispredicting a branch, then executing 300 instructions and having to throw them away once the branch is actually executed.