//template_cpu.cpp

 extern "C" void computeGold(float *, const unsigned int);

 void computeGold(float *idata, const unsigned int len)

 {

     const float f_len = static_cast<float>(len);

     for (unsigned int i = ; i < len; ++i)

         idata[i] *= f_len;

 }

 // template.cu

 #include <stdio.h>

 #include <cuda_runtime.h>

 #include "device_launch_parameters.h"

 #include <helper_functions.h>

 extern "C" void computeGold(float *, const unsigned int);

 __global__ void testKernel(float *g_idata, float *g_odata)

 {

     extern  __shared__  float sdata[];

     const unsigned int tid = threadIdx.x;

     sdata[tid] = g_idata[tid];

     __syncthreads();

     sdata[tid] = (float)blockDim.x * sdata[tid];

     __syncthreads();

     g_odata[tid] = sdata[tid];

 }

 int main()

 {

     printf("\n\tStart.\n");

     cudaSetDevice();

     StopWatchInterface *timer = ;

     sdkCreateTimer(&timer);

     sdkStartTimer(&timer);

     unsigned int num_threads = ;

     unsigned int mem_size = sizeof(float) * num_threads;

     float *h_idata, *h_odata, *d_idata, *d_odata;

     h_idata = (float *)malloc(mem_size);

     h_odata = (float *)malloc(mem_size);

     cudaMalloc((void **) &d_idata, mem_size);

     cudaMalloc((void **)&d_odata, mem_size);

     for (unsigned int i = ; i < num_threads; ++i)

         h_idata[i] = (float)i;

     cudaMemcpy(d_idata, h_idata, mem_size, cudaMemcpyHostToDevice);

     testKernel << < dim3(, , ), dim3(num_threads, , ), mem_size >> > (d_idata, d_odata);

     //getLastCudaError("Kernel execution failed");// 检查内核调用的报错结果

     cudaMemcpy(h_odata, d_odata, sizeof(float) * num_threads, cudaMemcpyDeviceToHost);

     cudaDeviceSynchronize();

     sdkStopTimer(&timer);

     printf("\n\tProcessing time: %f ms\n", sdkGetTimerValue(&timer));

     sdkDeleteTimer(&timer);

     computeGold(h_idata, num_threads);

     printf("\n\tFinish, return %s.\n", compareData(h_idata, h_odata, num_threads, 0.0f,0.0f) ? "Passed" : "Failed");

     free(h_idata);

     free(h_odata);

     cudaFree(d_idata);

     cudaFree(d_odata);

     getchar();

     return ;

 }

    Start.

    Processing time: 101.169357 ms

    Finish, return Passed.