Code improvements of the Perlin noise generator.

[mgold]

When the top of the file lists a chain of ports and adaptations three projects deep, there's bound to be some cruft that accumulates.

• Replace the cosine lookup table with Math.cos. This is simpler and does not affect the output or performance.
• Rename noise_fsc to scaled_cosine, and move it out of noise so it's not constantly redefined.
• Remove comments referencing the toxi bug tracker, from which this code was derived
• Remove a few comments asking if something was ok - I either fixed it or agree it's ok

Regarding the removal of the cosine lookup table, this seems to have been ported from toxi without much thought. Lookup tables aren't necessarily an optimization in JS, especially because of the memory overhead (SO thread). It also had a half-dozen constants, some of them poorly named. I was able to replace all of this without affecting either to output of the generator or performance on this test sketch:

sideLen = 400;
noiseScale = 0.02;

function setup() {
  createCanvas(sideLen, sideLen)
  noFill();
  noLoop();
  noiseSeed(42);
}


function draw() {
  for (i = 0; i < sideLen; i++){
    for (j = 0; j < sideLen; j++){
      stroke(noise(i*noiseScale, j*noiseScale)*255);
      point(i,j);
    }
  }
}

[shiffman]

Thanks @mgold for this contribution! I did the original porting and you are absolutely right that my goal was to just quickly get things in p5.js and working. Now's a great time to revisit for improvements and optimizations. @lmccart, this pull request looks good to me, I'll wait for you to merge in case you have any further comments / feedback.

(As an aside to this, these threads are relevant in terms of whether we want to break from the original Processing implementation with more "current" Perlin noise implementations.)

processing/processing#2549
processing/processing-docs#51

[lmccart]

beautiful, thank you!