#include <cuda_runtime.h>

#include <cstdio>

__global__ void increment(int* value) {
    *value += 1;
}

int main() {
    int host_value = 41;
    int* device_value = nullptr;

    if (cudaMalloc(&device_value, sizeof(int)) != cudaSuccess) {
        return 1;
    }
    if (cudaMemcpy(device_value, &host_value, sizeof(int), cudaMemcpyHostToDevice) != cudaSuccess) {
        cudaFree(device_value);
        return 2;
    }

    increment<<<1, 1>>>(device_value);
    if (cudaDeviceSynchronize() != cudaSuccess) {
        cudaFree(device_value);
        return 3;
    }
    if (cudaMemcpy(&host_value, device_value, sizeof(int), cudaMemcpyDeviceToHost) != cudaSuccess) {
        cudaFree(device_value);
        return 4;
    }

    cudaFree(device_value);
    std::printf("cuda_smoke_result=%d\n", host_value);
    return host_value == 42 ? 0 : 5;
}