Floating-point operations per second (FLOPS) is a common benchmark measurement for rating the speed of microprocessors. Floating-point operations include any operations that involve fractional numbers. Such operations, which take a lot longer to compute than integer operations, occur often in some applications.
The FLOPS measurement measures the speed of a floating-point unit (FPU). Most modern microprocessors include a FPU, which is a specialized part of the microprocessor responsible for executing floating-point operations. One of the most common benchmark tests used to measure FLOPS is called Linpack.
Many experts feel that FLOPS is not a relevant measurement because it fails to take into account factors such as the condition under which the microprocessor is running (e.g., heavy or light loads) and which exact operations are included as floating-point operations. For this reason, a consortium of vendors created the Standard Performance Evaluation Corporation (SPEC), which provides more meaningful benchmark values.
megaFLOPS (MFLOPS) = one million floating-point operations per second.
gigaFLOPS (GFLOPS) = one billion floating-point operations per second.
teraFLOPS (TFLOPS) = one trillion floating-point operations per second.
petaFLOPS = one quadrillion (million billion) floating-pointoperations per second.