>> And ticket locks do have nice real-time properties, as they get rid of
>> the possibility of starvation!
Is that a correct statement? Not an x86 guru but whatever instruction is used to atomically increment the 16bit number is that guaranteed to complete in hw? In other words consider a 16 way processor if all threads incremented simulataneously is there a guarantee all 16 threads increment (in whatever order) or is it possible that on some of them the thread can context switch out before execution. If such is true then it does not guarantee no starvation.