Instruction which computes a * b + c; usually in a minimal amount of time.  Very useful for efficiently computing DotProduct of two vectors, which is used in many convolution and filtering applications.  Often referred to as MAC (not to be confused with Media Access Control, which is what "MAC" stands for in "MAC address").

DigitalSignalProcessor''''''s frequently have performance quoted in MACs per second (or millions thereof), but like MIPS is not the best of metric; only really useful for comparing devices that are architecturally similar (if not identical).