An algorithm implementing a model whose results are known, and from the known results, the benchmarking of the device could be done. What is the currently known shortest circuit depth algorithm that has been used on a real device? An additional criterion is that it should be scalable.
As far as I know there isn't a "once size fits all" algorithm, rather random bench-marking is performed to have a statistical accuracy of the gates your computer has. There wouldn't be much point to make it just as short as possible since you would ideally try as many times as you can your gates to get an accuracy for your computer.