This is exported from the core library, which means you don't need to import it.
== two numbers must be of the same type, and have the same bit-pattern. As such, -0.0 == 0.0, but not identical -0.0 0.0.identical "tw" + "in" "twin".identical float.NAN float.NAN, but not identical float.NAN (float.from-bits float.NAN.bits + 1). This is unlike == where NaN's are never equal, so float.NAN == float.NAN is always false.
round-down 27 10 // => 20
round-down 7 2 // => 6
round-down -13 10 // => OUT_OF_RANGE error
round-up 17 10 // => 20
round-up 7 2 // => 8
round-up -13 10 // => OUT_OF_RANGE error