For what it's worth, I mostly[1] agree with you. You should consider should writing this up more formally as a challenge. Figure out some some of money you can

For what it's worth, I mostly[1] agree with you. You should consider should writing this up more formally as a challenge. Figure out some some of money you can afford to lose, and award it as a cash prize to anyone who can disprove you. show how measuring the execution time as often as you like lets you discern between one input buffer version and the other I think you can simplify this even further: is there any set of two inputs that can be distinguished from each other by measurement of execution time? One does not even need to identify which input is which --- merely show that there is some measurable property that is different for hashing A vs hashing B. [1] I think the weakness is the reference to a "reasonable C compiler". It seems likely to me that at least one of GCC/Clang/MSVC/ICC with some combination of legal flags will generate assembly that defeats the obvious intent of your algorithm. This would be a disappointing way to lose your bet.