Macro math 

math!() { /* proc-macro */ }

Changes all the math to fast float math. See fadd_fast.

This allows automatic vectorization of float math for the compiler, among other things.

This would lower a * b + c to fastadd(fastmul(a, b), c).