Floating-point unit

Acronym for Floating Point Unit, a specialized component or processor for arithmetic (and usually trigonometric) operations on numbers.

FPUs can be separate companion chips (such as the Intel 8087 for the main processor Intel 8086) or integrated on the same circuit as the processor itself.

A FPU has usually its own data registers and a subfamily of opcodes.

FPU instructions can be emulated by software in certain systems when the chip is not present.

See Floating Point Arithmetic, IEEE, NaN