WebGL and runtime performance in a functional paradigm

It’s different in every community. I’d recommend Code is the Easy Part and this post for info on this particular community.

My technical comment would be that compiler speed is a priority, especially for projects with 100k+ lines of code. So getting better runtime perf for P% of the community must be balanced against the compile time cost to the (100 - P)% of the community that does not benefit appreciably from better runtime perf. I do not think that hiding things behind flags is a full solution to this. More generally, there are no guarantees when working on projects like this, so think of your work as personal exploration. The perf numbers may be good, but the compile time cost, code quality, code complexity, working relationships, etc. etc. may mean that the code is not right for the project overall, even though the learnings are worth sharing.

Also, I was thinking more about the “see if there is only one reference” trick. The problem is that there may be only one direct reference, but many transitive references to the value, so this check is a bit trickier than I mentioned before. You’d have to be sure it was allocated locally to guarantee that there are no transitive references, making it a bit more limited.

Anyway, check out those links. Those are the best resources we have for this sort of thing!

1 Like