CUDA Array Sum with Reduction

mp4.c

// MP 4 Reduction
// Given a list (lst) of length n
// Output its sum = lst[0] + lst[1] + ... + lst[n-1];
// Due Tuesday, January 15, 2013 at 11:59 p.m. PST

#include    <wb.h>

#define BLOCK_SIZE 512 //@@ You can change this

#define wbCheck(stmt) do {                                 \
        cudaError_t err = stmt;                            \
        if (err != cudaSuccess) {                          \
            wbLog(ERROR, "Failed to run stmt ", #stmt);    \
            return -1;                                     \
        }                                                  \
    } while(0)

__global__ void total(float * input, float * output, int len) {
    //@@ Load a segment of the input vector into shared memory
    __shared__ float partialSum[2 * BLOCK_SIZE];
    unsigned int t = threadIdx.x, start = 2 * blockIdx.x * BLOCK_SIZE;
    if (start + t < len)
       partialSum[t] = input[start + t];
    else
       partialSum[t] = 0;
    if (start + BLOCK_SIZE + t < len)
       partialSum[BLOCK_SIZE + t] = input[start + BLOCK_SIZE + t];
    else
       partialSum[BLOCK_SIZE + t] = 0;
    //@@ Traverse the reduction tree
    for (unsigned int stride = BLOCK_SIZE; stride >= 1; stride >>= 1) {
       __syncthreads();
       if (t < stride)
          partialSum[t] += partialSum[t+stride];
    }
    //@@ Write the computed sum of the block to the output vector at the 
    //@@ correct index
    if (t == 0)
       output[blockIdx.x] = partialSum[0];
}

int main(int argc, char ** argv) {
    int ii;
    wbArg_t args;
    float * hostInput; // The input 1D list
    float * hostOutput; // The output list
    float * deviceInput;
    float * deviceOutput;
    int numInputElements; // number of elements in the input list
    int numOutputElements; // number of elements in the output list

    args = wbArg_read(argc, argv);

    wbTime_start(Generic, "Importing data and creating memory on host");
    hostInput = (float *) wbImport(wbArg_getInputFile(args, 0), &numInputElements);

    numOutputElements = numInputElements / (BLOCK_SIZE<<1);
    if (numInputElements % (BLOCK_SIZE<<1)) {
        numOutputElements++;
    }
    hostOutput = (float*) malloc(numOutputElements * sizeof(float));

    wbTime_stop(Generic, "Importing data and creating memory on host");

    wbLog(TRACE, "The number of input elements in the input is ", numInputElements);
    wbLog(TRACE, "The number of output elements in the input is ", numOutputElements);

    wbTime_start(GPU, "Allocating GPU memory.");
    //@@ Allocate GPU memory here
    cudaMalloc(&deviceInput, sizeof(float) * numInputElements);
    cudaMalloc(&deviceOutput, sizeof(float) * numOutputElements);

    wbTime_stop(GPU, "Allocating GPU memory.");

    wbTime_start(GPU, "Copying input memory to the GPU.");
    //@@ Copy memory to the GPU here
    cudaMemcpy(deviceInput, hostInput, sizeof(float) * numInputElements, cudaMemcpyHostToDevice);

    wbTime_stop(GPU, "Copying input memory to the GPU.");
    //@@ Initialize the grid and block dimensions here
    dim3 dimGrid(numOutputElements, 1, 1);
    dim3 dimBlock(BLOCK_SIZE, 1, 1);

    wbTime_start(Compute, "Performing CUDA computation");
    //@@ Launch the GPU Kernel here
    total<<<dimGrid, dimBlock>>>(deviceInput, deviceOutput, numInputElements);

    cudaDeviceSynchronize();
    wbTime_stop(Compute, "Performing CUDA computation");

    wbTime_start(Copy, "Copying output memory to the CPU");
    //@@ Copy the GPU memory back to the CPU here
    cudaMemcpy(hostOutput, deviceOutput, sizeof(float) * numOutputElements, cudaMemcpyDeviceToHost);

    wbTime_stop(Copy, "Copying output memory to the CPU");

    /********************************************************************
     * Reduce output vector on the host
     * NOTE: One could also perform the reduction of the output vector
     * recursively and support any size input. For simplicity, we do not
     * require that for this lab.
     ********************************************************************/
    for (ii = 1; ii < numOutputElements; ii++) {
        hostOutput[0] += hostOutput[ii];
    }

    wbTime_start(GPU, "Freeing GPU Memory");
    //@@ Free the GPU memory here
    cudaFree(deviceInput);
    cudaFree(deviceOutput);

    wbTime_stop(GPU, "Freeing GPU Memory");

    wbSolution(args, hostOutput, 1);

    free(hostInput);
    free(hostOutput);

    return 0;
}

Works but is 10000 times slower than cpu. I seldom get a GPU implementation that beats CPU especially when I make the CPU parallel. I don't get it!