Comment made by: favila
I'll have time to do this in about a week. The implementation is straightforward (basically use xor 0 everywhere). The goal is correctness, but I expect performance to be as good as or better than it is now on most platforms. I'm not sure if advanced mode will drop intermediate truncations or what impact this has on performance.
Some higher-level numeric analysis using the asm.js type system is possible but I doubt it's worth it.