root/test/correctness/gpu_vectorized_shared_memory.cpp

DEFINITIONS

1. main

#include "Halide.h"
#include <stdio.h>

using namespace Halide;

int main(int argc, char **argv) {
    // This test is only relevant for opencl
    if (get_jit_target_from_environment().has_feature(Target::OpenCL)) {

        Func f, g, h;
        Var x;

        f(x) = x;
        g(x) = f(x-1) + f(x+1);
        RDom r(0, 10);
        g(x) += sum(r);
        h(x) = g(x-1) + g(x+1);

        f.compute_root();

        Var xo, xi;
        h.split(x, xo, xi, 16).vectorize(xi, 4).gpu_threads(xi).gpu_blocks(xo);
        g.compute_at(h, xo);
        g.split(x, xo, xi, 4).gpu_threads(xo).vectorize(xi);
        g.update().split(x, xo, xi, 4).gpu_threads(xo).vectorize(xi);

        Buffer<int> out = h.realize(512);

        for (int x = 0; x < out.width(); x++) {
            int correct = 4*x + 90;
            if (out(x) != correct) {
                printf("out(%d) = %d instead of %d\n",
                       x, out(x), correct);
                return -1;
            }
        }

    } else {
        printf("This test is only relevant for OpenCL targets\n");
        return 0;
    }

    printf("Success!\n");
    return 0;

}