A SWP specification for sequential image processing algorithms

  • Authors:
  • Wensheng Tang;Shaogang Wang;Dan Wu;Wangqiu Kuang

  • Affiliations:
  • Computer School, National University of Defense Technology, Changsha, Hunan, China;Computer School, National University of Defense Technology, Changsha, Hunan, China;Computer School, National University of Defense Technology, Changsha, Hunan, China;Computer School, National University of Defense Technology, Changsha, Hunan, China

  • Venue:
  • ACSAC'07 Proceedings of the 12th Asia-Pacific conference on Advances in Computer Systems Architecture
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

To retarget sequential image processing algorithm written in sequential languages (e.g. C) to processors with multimedia extensions and multimedia-specific embedded microprocessors is an open issue. Image processing algorithms are inherently with sub-word parallelism (SWP) But current compilers are not able to exploit it by locating SWP within a basic block. This paper proposes a program representation and pattern-matching approach for generating an explicit SWP specification from sequential source code. The representation is based on an extension of the multidimensional synchronous dataflow (MDSDF) model of computation. For both the compiler and source programs should not be user-modified, we extend compiler's functionality by adding a specialized pattern-library. After data-flow and control-flow analysis with pattern matching, the generated SWFG (Sub-Word Flow Graph) can be used as an intermediate representation for the next step of compiler for SWP instruction selection and code generation.