The media extensions to the Intel Architecture (IA) instruction set include single-instruction, multiple-data (SIMD) instructions. Streaming SIMD Extensions 2 (SSE2) instructions extend SIMD for the Intel NetBurst® microarchitecture with 144 new instructions. This paper describes an SSE2 implementation to calculate the absolute difference between two 16x16 blocks of pixels. This paper also compares the performance gains of that solution to non-SSE2 implementations. This SSE2 solution can be an integral part of a motion-estimation kernel.
Determine whether the target system is based on the Pentium® M processor. Before enabling optimizations that are specific to the Pentium M processor, an application should ensure that the machine it is running on is based on a Pentium M processor.