• DocumentCode
    2393898
  • Title

    Compiler driven data cache prefetching for high performance computers

  • Author

    Chi, Chi-Hung ; Fang, Kam-Kong

  • Author_Institution
    Dept. of Comput. Sci., Chinese Univ. of Hong Kong, Shatin, Hong Kong
  • fYear
    1994
  • fDate
    22-26 Aug 1994
  • Firstpage
    274
  • Abstract
    With advances in compiler optimization and program flow analysis, software assisted cache prefetching schemes using PREFETCH instructions are now possible. Although data can be prefetched accurately into the cache, the runtime overhead associated with these schemes often limits their practical use. We propose a new prefetching scheme, called the Embedded Data Prefetching (EDP), to prefetch array references with constant strides accurately. The basic idea of the EDP scheme is to store constant stride values of array references to be prefetched into a register array, called the stride register file. For each array reference r with constant stride s, the name of the stride register containing s will be embedded into the corresponding LOAD or STORE instruction for r. Then, prefetch hardware will use this embedded prefetching hint and its corresponding value in the stride register for accurate data prefetching. Compared to current software assisted data prefetching schemes, the EDP scheme has much lower runtime overhead without sacrificing prefetching accuracy. Our result showed that the EDP scheme is particularly suitable for computing intensive scientific applications where cache misses are mainly due to array references with constant strides and they can be prefetched very accurately by this EDP scheme
  • Keywords
    cache storage; data flow analysis; data handling; optimisation; program compilers; Embedded Data Prefetching; PREFETCH instructions; array references; cache misses; compiler driven data cache prefetching; compiler optimization; constant strides; embedded prefetching; high performance computers; intensive scientific applications; prefetch hardware; program flow analysis; register array; runtime overhead; stride register; stride register file; Application software; Computer aided instruction; Computer applications; Computer science; Hardware; High performance computing; Optimizing compilers; Prefetching; Program processors; Runtime;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    TENCON '94. IEEE Region 10's Ninth Annual International Conference. Theme: Frontiers of Computer Technology. Proceedings of 1994
  • Print_ISBN
    0-7803-1862-5
  • Type

    conf

  • DOI
    10.1109/TENCON.1994.369294
  • Filename
    369294