Calculator benchmark

[robve]

(08-15-2021 11:19 PM)Maximilian Hohmann Wrote:  I would do it with a stopwatch. Count the seconds from turning on the calculator to getting the last result.

I'm afraid this is going nowhere. There are serious problems with this. When properly prepared and experienced with the calculator at hand, each keystroke should take a just about a fixed amount of time by a human to operate and register for this simple benchmark. If keystrokes are not quick or easy to make, because of the way the calculator works, then you are back to the problem you stated: "not how fast the calculator works internally", but that is exactly what you could end up timing. You can't have it both ways. The other problem is that time, as a quantitive measure of performance of a calculator operated by a human, is statistically and subjectively noisy. You will first need to establish a fair measure of timing to begin with. You can either take the fastest time among many users of the calculator, but then you're back at my first point, or you can determine the mean and deviation of the time to operate the calculator using a little army of human testers. The latter follows the traditional scientific method to ensure reproducibility of the experiment.

It seems to me that all of this is totally out of scope for a simple fun benchmark experiment. Therefore, stick to the number of keystrokes as unbiased and reproducible.

For what counts as a straightforward method, I will only speak for my examples and improvements, which are very basic and easy programs with executions that do not use any specialized or uncommon features of a calculator. For example, the PC-E500 program and execution are typical for using this BASIC calculator, requiring nothing special.

Furthermore, unnecessary keystrokes should not count. For example, it is incorrect to rank Voyage 200 at 175 keystrokes when a normal straightforward method to solve this problem requires 121 keystrokes, simply by avoiding unnecessary keystrokes.

- Rob