/**
 * Use of SSE vector units
 */
f32 dp_sse_vec(f32 *x, f32 *y, u32 size) {
	// for example use xmm0 to store v_sum
    f32 v_sum[4];
    // i must be defined here because it will be used
    // by the for loop and then the while loop
    u32 i;
    
    // use xorps xmm0, xmm0
    v_sum[0] = v_sum[1] = v_sum[2] = v_sum[3] = 0.0;
   
    // loop unrolling by a factor of 4
    // use xmm1 to store x[i:i+3]
    for (i = 0; i < (size & ~3); i += 4) {
        v_sum[0] += x[i + 0] * y[i + 0];
        v_sum[1] += x[i + 1] * y[i + 1];
        v_sum[2] += x[i + 2] * y[i + 2];
        v_sum[3] += x[i + 3] * y[i + 3];
    }
    
    // sum of partial sums
    float sum = v_sum[0] + v_sum[1] + v_sum[2] + v_sum[3];
    
    // last iterations
    while (i < size) {
    	sum += x[i] * y[i];
    	++i;
    }
   
    return sum;
}