Introduction to CUDA C

Dot Product
� Parallel threads have no problem computing the pairwise products:
a b
a 0
a 1
a 2
a 3
*
*
*
*
b 0
b 1
b 2
b 3
� So we can start a dot product CUDA kernel by doing just that:
__global__ void dot( int *a, int *b, int *c ) {
// Each thread computes a pairwise product
int temp = a[threadIdx.x] * b[threadIdx.x];
+