2 Replies Latest reply on Jun 20, 2010 8:16 PM by kbrafford

    Optimizing a kernel

    kbrafford

      I am using the stream analyzer for the first time.  Can someone explain (or point me to a primer on) how I interpret this table?  I'm using a 5870 and I'd like to get some more umph out of my code if I can, plus I'd like to learn the proper process that one uses in general.

       

       

      NameGPRScratch RegMinMaxAvgALUFetchWriteEst CyclesALU:FetchBottleNeck%s\ClockThroughput
      Radeon HD 489019044.6546.8171.93265422171.937.59ALU Ops0.0979 M Threads\Sec
      Radeon HD 477019055.75683.5214.91265422214.917.59ALU Ops0.0756 M Threads\Sec
      Radeon HD 487019044.6546.8171.93265422171.937.59ALU Ops0.0970 M Threads\Sec
      Radeon HD 467016661.88779.25198.393159218198.397.25ALU Ops0.0430 M Threads\Sec
      Radeon HD 45501902232734707.71265422707.7119.82ALU Ops0.017 M Threads\Sec
      Radeon HD 587021042468152.94262822152.946.75ALU Ops0.21178 M Threads\Sec
      Radeon HD 577021042468152.94262822152.946.75ALU Ops0.189 M Threads\Sec
      Radeon HD 567021084936257.52262822257.5218.03ALU Ops0.0324 M Threads\Sec
      Radeon HD 54502102102340571.95262822571.9549.38ALU Ops0.015 M Threads\Sec
      FireStream 925019044.6546.8171.93265422171.937.59ALU Ops0.0958 M Threads\Sec
      FireStream 927019044.6546.8171.93265422171.937.59ALU Ops0.0970 M Threads\Sec