@delfrrr can you also fix the misleading statement in the readme about 6x faster than JS, or at least explain how you got this result?

@mourner will do soon, need to dig into results;
I made current statements based on results from this script https://github.com/delfrrr/delaunator-cpp/blob/master/generate-reference-triangles/index.js

I agree that they may be not precise; also note the version I was testing against;

@delfrrr the script doesn't have any warmup, which is especially important for JS because it's JIT-optimized. It might have also measured a dataset that's too small.

I modified cpp and js benchmark to make it more similar;
for JS benchmark I removed part where points are transformed to typed array, so results are even slightly better for JS. Also added more samples for CPP

I think there is ~30ms overhead in JS version which makes much slower on small number of points. On 1M points difference is ~10%; since time is not growing linear with number of points comparison in percents is not relevant.

@mourner btw, what is estimated complexity of algorithm itself?

@delfrrr it should be O(n log n) in average. The 30ms JS "overhead" is likely just the JIT warmup — subsequent runs on the same amount of points should be much faster.