megaFLOPS

megaFLOPS, abbreviated MFLOPS, stands for million floating-point operations per second. It is a unit of computer performance used to quantify the rate at which a processor can carry out floating-point arithmetic, typically counting operations such as additions, multiplications, and other floating-point calculations.

In practice, MFLOPS can refer to different floating-point precisions, often single- or double-precision, and conventions may vary on how certain operations are counted. For example, a fused multiply-add (FMA) instruction can be treated as one or two floating-point operations depending on the benchmark. Because of this, MFLOPS values should be interpreted in the context of the specific benchmark and precision used.

MFLOPS is a legacy metric that was widely used during the early era of digital computers and mid- to late-20th-century supercomputing. As hardware matured, the standard units shifted toward higher magnitudes such as GFLOPS (billion FLOPS) and TFLOPS (trillion FLOPS), reflecting the much larger computational capabilities of modern processors.

Measuring MFLOPS typically involves running a benchmark that executes a known total number of floating-point operations and records the elapsed time. The performance is then computed as total FLOPs divided by the execution time, yielding MFLOPS. Despite its historical usefulness, MFLOPS provides a narrow view of performance and does not capture memory bandwidth, latency, or real-world efficiency.

See also: FLOPS, GFLOPS, TFLOPS, floating-point arithmetic, benchmarking.