6 Replies Latest reply on Mar 28, 2011 10:36 AM by Nil

    Experiment: Different runtime for cores running same code




      I was looking at placement of core and its effect on memory bound computation, the experiment consists for loop incrementing value of array elements (initialized to one). I used three different size for array (8192, 131072, 2097152). Graphs for result are attached.


      One thing I noticed in graphs is that runtime is high for cores which are in middle (1st and 3rd row and 2nd 3rd column). The question is:  it is the case because of the distance from core to MC or is there other thing playing part here.


      Any suggestion or explanation welcome.