accumulate.br

用于GPU通用计算的编程语言BrookGPU 0.4

// accumulate.br
//
// This test repeatedly computes b += a for two different streams.  The
// goal is to validate that read-modify-write works with kernels (i.e. the
// same stream can be passed as an input and an output).

#include <stdio.h>

static void callKernel(float a<>, float b<>);

kernel void
sum(float a<>, float b<>, out float c<>)
{
   c = a + b;
}

static void
callKernel(float a<>, float b<>)
{
   sum(a, b, b);
}

int
main(int argc, char *argv[])
{
   float input_a1[100], input_b1[100], output_b1[100];
   float output_b2[10][10];
   float a1<100>, b1<100>;
   float a2<10,10>, b2<10,10>;
   int i, j;

   for (i = 0; i < 10; i++ ) {
      for (j = 0; j < 10; j++ ) {
         input_a1[10*i+j] = 0.125f*(float)(10*i+j);
         input_b1[10*i+j] = (float)0;
      }
   }
   streamRead(a1, input_a1); streamRead(a2, input_a1);
   streamRead(b1, input_b1); streamRead(b2, input_b1);

   for (i = 0; i < 12; i++ ) {
      callKernel(a1, b1);
      callKernel(a2, b2);
   }

   streamWrite(b1, output_b1);
   streamWrite(b2, output_b2);

   for (i = 0; i < 10; i++) {
      for (j = 0; j < 10; j++)
         printf( "%6.2f ", output_b1[10*i+j] );
      printf("\n");
   }

   for( i = 0; i < 10; i++ ) {
      for( j = 0; j < 10; j++ )
         printf( "%6.2f ", output_b2[i][j] );
      printf("\n");
   }

   return 0;
}