Multi-media extensions in super-pipelined micro-architectures. A new case for SIMD processing?

Abstract

General purpose microprocessors have long been considered a computing platform unsuited to image processing and vision tasks. The so-called Von-Neumann paradigm and the associated memory bottleneck have motivated the research into various forms of parallel processing and of special processors for vision. The SIMD approach, adopted in massively parallel processors, has been introduced in a minimal format in the multimedia extensions to instruction set architectures of standard microprocessors. This paper examines the characteristics of SIMD processing that have been mapped into these extensions.