As a function, it rounds its argument to the nearest IEEE 754 single precision (i.e. IEEE754-2008 binary32) number.
Subnormal numbers are supported as well as standard numbers: it is the real
rounding described in the standard.
As a constant, it symbolizes the single precision format. It is used in
contexts when a precision format is necessary, e.g. in the commands
round and roundcoefficients. In is not supported for implementpoly.
See the corresponding help pages for examples.