SIMD in the 90s: Programming Intel's Pentium MMX
Summary
An historical overview of MMX and SIMD on the Pentium. It explains MMX registers, the concept of data packing, and how 57 instructions enabled parallel operations on a single 64 bit register. The article traces the evolution from MMX to SSE and SSE2, compares MMX to 3DNow, describes CPUID based detection, and notes practical use cases such as image processing and audio mixing. It also discusses limitations such as the mutual exclusivity with x87 and how developers ported between MMX and non MMX code.