Global memory bandwidth profiling

Press the button to get a new bandwidth value for copying 2048×2048 float matrices with (32×8) thread blocks.

__kernel void kMatrixCopy( global float* A, global float* C ){
  int x = get_global_id(0) + (get_global_id(1) << 11);
  C[x] = A[x];
}

WebCL Demos     updated 17 Nov 2011