(computing) An extension to the x86 CPU instruction set architecture, providing direct hardware support for advanced floating-point arithmetic operations (rather than requiring them to be emulated in software).