PSLab5Test4

#include <stdio.h>
#define SIZE 100
#define BLOCKSIZE 100


void HANDLE_ERROR(cudaError_t call)
{
    cudaError_t ret = call;
    switch (ret)
    {
    case cudaSuccess:
        break;
    case cudaErrorInvalidValue:
        printf("ERROR: InvalidValue:%i.\n", __LINE__);
        exit(-1);
        break;
    case cudaErrorInvalidMemcpyDirection:
        printf("ERROR:Invalid memcpy direction:%i.\n", __LINE__);
        exit(-1);
        break;
    default:
        printf("ERROR>line:%i.%d' ‘ %s\n", __LINE__, ret, cudaGetErrorString(ret));
        exit(-1);
        break;
    }
}

__device__ int pomnozi(int a, int b)
{
    return a*b;
}

__global__ void compute(int *d_a, int* d_b, int* d_c)
{
    d_c[threadIdx.x] = 0;
    for (int i = 0; i<SIZE; i++) {
        int ela = d_a[threadIdx.x * i];
        int elb = d_b[i];
        d_c[threadIdx.x] += pomnozi(ela, elb);
    }
}

__host__ void outer_compute(int** h_a, int* h_b, int* h_c)
{
    int *d_a, *d_b, *d_c, i;
    HANDLE_ERROR(cudaMalloc((void **)& d_a, SIZE * SIZE * sizeof(int)));
    HANDLE_ERROR(cudaMalloc((void **)& d_b, SIZE * sizeof(int)));
    HANDLE_ERROR(cudaMalloc((void **)& d_c, BLOCKSIZE * sizeof(int)));

    //kopiranje nizova iz CPU u GPU
    for(i = 0; i < SIZE * SIZE; i = i + SIZE)
    HANDLE_ERROR(cudaMemcpy(d_a[i], h_a[i], SIZE * sizeof(int), cudaMemcpyHostToDevice));
    HANDLE_ERROR(cudaMemcpy(d_b, h_b, SIZE * sizeof(int), cudaMemcpyHostToDevice));

    compute<<<1, BLOCKSIZE >>>(d_a, d_b, d_c);

    //sinhronizovanje i kopiranje niza iz GPU u CPU
    cudaThreadSynchronize();
    HANDLE_ERROR(cudaMemcpy(h_c, d_c, BLOCKSIZE * sizeof(int), cudaMemcpyDeviceToHost));

    cudaFree(d_a);
    cudaFree(d_b);
    cudaFree(d_c);
}


int main(int argc, char **argv)
{
    int i, j;
    int **a, b[SIZE], c_array[BLOCKSIZE];
    int sumCuda = 0, sumClassic=0;

    a = (**int) malloc(SIZE * sizeof(int*));
    for (i = 0; i < 100; i++) {
        a[i] = (*int) malloc(SIZE * sizeof(int));
    }

    //inicijalizacija
    for (i = 0; i < SIZE; ++i) {
        b[i] = i;
        for (j = 0; j < SIZE; ++j) {
            a[i][j] = i * SIZE + j;
        }
    }

    //pozivanje host funkcije za CUDA i
    outer_compute(a, b, c_array);
    //Sabiranje rezultata dobijenih iz niti
    for (i = 0; i < SIZE; i++) {
        printf("Rezultat paralelni %d \n", c_array[i]);
        c_array[i] = 0;
    }


    //sekvencijalan kod
    for (i = 0; i<SIZE; i++) {
        for (j = 0; j < SIZE; ++j)
            c_array[i] += a[i][j] * b[j];
    }


    for (i = 0; i < SIZE; i++) {
        printf("Rezultat sekvencijalni %d \n", c_array[i]);
        c_array[i] = 0;
    }

    //printf("Razlika izmedju dva rezultata je %d",sumClassic-sumCuda);
}