Brook+ GPU

Brook+ is a C stream programming model for numerical computations on the AMD/ATI Radeon graphics processing units (GPU).

A simplified translation of the following example parallel-for loop is given below.
Grid1 *g = new Grid1(0, n+1);
Grid1IteratorSub it(1, n, g);
DistArray x(g), y(g);
...
float e = 0;
ForEach(int i, it,
   x(i) += ( y(i+1) + y(i-1) )*.5;
   e += sqr( y(i) ); )
...


kernel code:
kernel void sub1(float x[], float y[],
  out float xout<>, out float eout<>)
{
  float i = 1 + indexof(xout).x;
  xout = x[i] + (y[i+1] + y[i-1])*.5f;
  eout = y[i] * y[i];
}

reduce void sub2(float y<>, reduce float e<>)
{
  e += y;
}

main code:
unsigned int n = 8190;
unsigned int n2 = n + 2;
float *x = (float*)malloc(sizeof(float) * n2);
float *y = (float*)malloc(sizeof(float) * n2);
float e;
...
{
  float xStream<n>;
  float xoutStream<n>;
  float yStream<n2>;
  float einStream<n2>;
  float eStream<1>;
  streamRead(xStream, x+1);
  streamRead(yStream, y);
  sub1(xStream, yStream, xoutStream, einStream);
  streamWrite(xoutStream, x+1);
  sub2(einStream, eStream);
  streamWrite(eStream, &e);
}
...
free(y);
free(x);

Arrays larger than the maximum of 8192 in one dimension require some code modifications. Note the limited programming features for the kernel code.