Unreadable memory usage and runtime graphs

[SebastianMestre]

Good morning!

The graphs in the README are notreadable. All that they show is that time and space usage are quadratic on the side length, which is obvious, since the amount of data is quadratic on the side length.

I suggest changing the x-axis to be the number of pixels. Then, the relationship between the amount of input pixel data and time/space usage will be clearer. If the graph is linear, it can be used to get a sense of throughput (the reciprocal of the slope is pixels per second) and memory overhead (the slope would be bytes per pixel).

[big-nacho]

which is obvious, since the amount of data is quadratic on the side length

Unless I misunderstood, I don't think this is true.
Suppose the runtime complexity on the number of pixels is given by $T$ and the runtime complexity on the side length by $S$. Then it holds that $S(n) = T(n^2)$, but that doesn't mean the runtime complexity on the side length is quadratic (suppose $T(n) = e^n$, for example).

In any case the purpose of the graphs was to give a rough overview of what can be expected for typical image resolutions, and I think side length is how most people think about those, usually. They weren't meant to show any "proper" complexity analysis.

I'll keep this open as I plan to update the README soon and I do agree that the graphs could be more helpful, regardless.

[SebastianMestre]

I don't wanna get in your hair too much here, but I think something that lets you grasp the pixels per second and bytes per pixel at a glance would be better for programmer-type users.

I agree that basing it on side lengths is more useful for end users. Although a table might be more so than a graph?

Would you consider having both? Maybe the side length one in a more visible spot, then the pixel count one tucked away further below?

[big-nacho]

I don't wanna get in your hair too much here

Not at all, mate. Thanks for the feedback.

I would like to wait until I do some proper work on memory usage, though. At the moment I feel the takeaway will be "this thing uses an insane amount of memory" no matter the format. Once that's done (it's pretty much my first priority), then I'd be happy to update the graphs 😄