At the time I wrote those, I was actually surprised because I didn't see much improvement using emcc-raycast. In both implementations, you can see that I'm allocating objects/structs all over the place, so that probably brings down the performance.