even without likely/unlikely, branch prediction still occurs in branch-prediction supporting processors. the kernel optimization just optimizes for the first ca

even without likely/unlikely, branch prediction still occurs in branch-prediction supporting processors. the kernel optimization just optimizes for the first case. once the branch prediction state machine as aligned itself towards the more likely branch, accurate branch prediction will occur for the resulting calls (typically iterations or something like that)