• DocumentCode
    3410582
  • Title

    Pre-decoding mechanism for superscalar architecture

  • Author

    Minagawa, Kenji ; Saito, Mitsuo ; Aikawa, Takeshi

  • Author_Institution
    Toshiba Res. & Dev. Center, Kawasaki, Japan
  • fYear
    1991
  • fDate
    9-10 May 1991
  • Firstpage
    21
  • Abstract
    A predecoding mechanism for superscalar processors that is necessary to issue instructions to plural functional units is described. Using a predecoding mechanism, a processor can issue instructions without increasing the cycle time or the branch penalty. The predecoder decides which functional unit should execute an instruction during a cache miss and refill. A special tag, the call predecode tag, is generated by the predecoder. This tag is saved with instructions to an instruction cache memory. The processor fetches instructions with the predecode tag and issues instructions controlled by the predecode tag. Using this mechanism, fast cycle times, comparable to those of RISC processors, are attained
  • Keywords
    parallel architectures; call predecode tag; fast cycle times; predecoding mechanism; superscalar processors; Cache memory; Decoding; Degradation; Delay; Pipelines; Process design; Reduced instruction set computing; Registers; Research and development; VLIW;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Communications, Computers and Signal Processing, 1991., IEEE Pacific Rim Conference on
  • Conference_Location
    Victoria, BC
  • Print_ISBN
    0-87942-638-1
  • Type

    conf

  • DOI
    10.1109/PACRIM.1991.160671
  • Filename
    160671