1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
|
$void __global__ TimesTwoKernel(double *A,
$ double *B,
$ int const N)
${
$ /* Calculate the global linear index, assuming a 1-d grid. */
$ int const i = blockDim.x * blockIdx.x + threadIdx.x;
$ if (i < N) {
$ B[i] = 2.0 * A[i];
$ }
$}
$void TimesTwo(double *A,
$ double *B,
$ int const N)
${
$ int const threadsPerBlock = 256;
$ int blocksPerGrid = (N + threadsPerBlock - 1) / threadsPerBlock;
$ TimesTwoKernel<<<blocksPerGrid, threadsPerBlock>>>(A, B, N);
$}
@function result = timestwo(a)
n = numel(a)
# TimesTwo(gpu double[] a, gpu output double[n] result, int n);
end
|
