Rotation matrix may be computed just once

[ednl]

This might speed up the simulation, but probably not enough to have an acceptable frame rate in real time. So I guess you would still need to make a time lapse. In that regard, I'm not sure this is worth the trouble.

What the code does now is increment the rotation angle every loop by 1 degree, construct a new rotation matrix and use that matrix only once. Alternatively, you could construct a rotation matrix just once, before the draw loop, with an angle of 1 degree. Then use that matrix every time to transform the current coordinates. This will have the same effect, cumulatively. So instead of using sin and cos every loop and do the matrix multiplication, you have only the multiplication left.

One slight complication is that matrix multiplication p' = R.p can't be done in-place, because the old values would be overwritten when you still need them. So in BASIC, this probably requires juggling with array indices where you transform p(0..7) into p(8..15) on one loop, and p(8..15) into p(0..7) on the next loop.

I hope this is clear and that it will be helpful. Cheers!

[ednl]

I made an example in P5js. In the sketch code, I cheated by making the points p and rotation matrix R multilevel arrays, which BASIC doesn't have, I think? So that would require some index juggling.

[shiffman]

Thanks so much for this, glad to be able to discuss during yesterday's live stream! If you would like to submit a pull request with an additional "optimized" cube.bas file that works for me! Or even a README.md file with an explanation! Or this issue can remain as a nice reference!