@UK0810AQ2 You've asked for some flamegraphs...

The first unoptimised version: Reflection dominates.

The rendered itself was quite respectable, taking 2-3 seconds to render a 640x480 scene, but actually converting the result to a PPM file (using a StringBuilder) was another 8-9 seconds.

After taking care of that, rendering now dominates - but a significant chunk of that was actually setting a pixel in the canvas array, because I was using the high-level set-selection! of core.matrix.

Replacing that with just setting 3 r g b values pushes that down even further. Now I see two instance of sort-by taking a long time to sort just a handful of hits so that's the next place to optimise. Everything else seems to bottom out more or less to the matrix operations. Perhaps allocating points/vectors could be faster as that happens a lot. Perhaps using a macro or something would make sense.

That's all for now 🙂 This is addicting. Next chapter is about adding shadows.

This is awesome 😄

and trimming down performance is addictive 😄

you can zoom in on each of the stacks. Maybe if you do it without mapcat you'll gain a huge speedup

I'd like som context, @U7PBP4UVA! 😃

@U0ETXRFEW https://clojurians.slack.com/archives/CBJ5CGE0G/p1644852220997739

Awesome.

It's nice to coding to a spec. The book gives you some test cases, then guides on you on how to implement the algorithms. It's actually quite relaxing.

I did most of the work while recovering from a cold.

very cool indeed!

I was sure core matrix was already under org.clojure. Did you model the tuple as array?

Yes, an array of 4 elements. Makes sense for all the following matrix multiplications. The book actually gives you guidance for all of this.