template <int BLOCK_SIZE> __global__ void sumKernelStr2(float *c, float*a) {
    unsigned int tid = 2*threadIdx.x;
    unsigned int i = blockIdx.x * 2*blockDim.x + 2*threadIdx.x;


    for (unsigned int odstep = 1; odstep < 2*blockDim.x; odstep *= 2) {
        int index = odstep*tid;
        if (index < 2*blockDim.x) {
            a[index] += a[index + odstep];
        }
    }

    if (tid == 0) c[blockIdx.x] = a[0];
}

template <int BLOCK_SIZE> __global__ void sumKernelStr3(float *c, float *a) {
    unsigned int tid = threadIdx.x;
    unsigned int i = blockIdx.x * 2*blockDim.x + threadIdx.x;

    for (unsigned int odstep = blockDim.x; odstep > 0; odstep /= 2) {
        if (tid < odstep) a[tid] += a[tid + odstep];
    }
    if (tid == 0) c[blockIdx.x] = sdata[0];
}