• DocumentCode
    2430429
  • Title

    Software de-pipelining technique

  • Author

    Su, Bogong ; Wang, Jian ; Hu, Erh-Wen ; Manzano, Joseph

  • Author_Institution
    William Paterson Univ., Wayne, NJ, USA
  • fYear
    2004
  • fDate
    15-16 Sept. 2004
  • Firstpage
    7
  • Lastpage
    16
  • Abstract
    Software pipelining is an optimization technique used to speed up loop execution. It is widely implemented in optimizing compilers for VLIW and superscalar processors that support instruction level parallelism. Software de-pipelining is the reverse of software pipelining; it restores the assembly code of a software-pipelined loop back to its semantically equivalent sequential form. Due to the non-sequential nature of the often optimized assembly code, it is very difficult to gain insight into the meaning of the code. Consequently, the task of de-pipelining the code of a software-pipelined loop is very complex and challenging. We present in This work our de-pipelining algorithm with a formal description, proof, and a set of working examples. Experiments with loops taken from some practical DSP programs are conducted on popular VLIW digital signal processors to verify the algorithm. Some applications of software de-pipelining are discussed.
  • Keywords
    instruction sets; optimising compilers; pipeline processing; program assemblers; program control structures; DSP programs; VLIW digital signal processors; assembly code restoration; formal description; instruction level parallelism; loop execution; optimization technique; optimizing compilers; semantically equivalent sequential form; software de-pipelining technique; software pipelining; software-pipelined loop back; superscalar processors; Application software; Assembly; Delay; Digital signal processing; Digital signal processors; Instruments; Optimizing compilers; Pipeline processing; Signal processing algorithms; VLIW;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Source Code Analysis and Manipulation, 2004. Fourth IEEE International Workshop on
  • Print_ISBN
    0-7695-2144-4
  • Type

    conf

  • DOI
    10.1109/SCAM.2004.20
  • Filename
    1386154