CUDA + OpenCV：此应用程序无法正常启动（0xc000007b）

Question

当我在windows7 x64上运行vs2008中的程序时，使用cuda4.1和opencv2.3.it会出现“应用程序无法正常启动（0xc000007b）....”，没有编译和链接错误，它只发生在运行时，它让我疯狂。任何帮助将是欣赏！谢谢！
这是我的代码：

#include <cutil_math.h>
#include <cutil_inline.h>
#include <highgui.h>
#include <cv.h>
#include <cxcore.h>
#include <book.h>
#include <cpu_bitmap.h>

#define ThreadX  16
#define ThreadY  16  

texture<float, 2, cudaReadModeElementType> texRefSource;
texture<float, 2, cudaReadModeElementType> texRefTarget;

__constant__ float c_trans[9];

__global__ void transformKernel(float* dev_bitmap,int width, int height)
{
    //计算拾取坐标
    unsigned int x = blockIdx.x * blockDim.x + threadIdx.x;
    unsigned int y = blockIdx.y * blockDim.y + threadIdx.y;

    float3 v=make_float3(x,y,1);
    float3 r0=make_float3(c_trans[0],c_trans[1],c_trans[2]);
        float3 r1=make_float3(c_trans[3],c_trans[4],c_trans[5]);
        float3 r2=make_float3(c_trans[6],c_trans[7],c_trans[8]);

    float tz=dot(r2,v);
        float tx=dot(r0,v)/tz;
    float ty=dot(r1,v)/tz;
    dev_bitmap[y * width + x] = tex2D(texRefTarget, tx, ty);
}

int main()
{  
    // load and set source file
    IplImage* source=cvLoadImage("C:\\1368.bmp"); 
    IplImage* sourcegray =cvCreateImage(cvGetSize(source),source->depth,1);
    cvCvtColor(source,sourcegray,CV_BGR2GRAY);//image1图像灰度化

    int w1 = sourcegray->width;
    int h1 = sourcegray->height;
    int size1=w1*h1;

    cudaArray* SourcecuArray;
    cudaChannelFormatDesc channelDesc1 = cudaCreateChannelDesc(32, 0, 0,                 0,cudaChannelFormatKindFloat);
    cudaMallocArray(&SourcecuArray, &channelDesc1, w1, h1);
    cudaMemcpyToArray(SourcecuArray, 0, 0, sourcegray->imageData,size1,cudaMemcpyHostToDevice);

    texRefSource.addressMode[0] = cudaAddressModeWrap; //循环寻址方式
    texRefSource.addressMode[1] = cudaAddressModeWrap;
    texRefSource.filterMode = cudaFilterModeLinear;

    cudaBindTextureToArray(texRefSource, SourcecuArray, channelDesc1);

    // load and set target file
    IplImage* target=cvLoadImage("C:\\1369.bmp");
    IplImage* targetgray =cvCreateImage(cvGetSize(target),target->depth,1);
    cvCvtColor(target,targetgray,CV_BGR2GRAY);//image1图像灰度化
    cvNamedWindow("before", CV_WINDOW_AUTOSIZE );
    cvShowImage("before", targetgray);

    int w2 = targetgray->width;
    int h2 = targetgray->height;
    int size2=w2*h2;

    cudaArray* TargetcuArray;
    cudaChannelFormatDesc channelDesc2 = cudaCreateChannelDesc(32, 0, 0, 0,cudaChannelFormatKindFloat);
    cudaMallocArray(&TargetcuArray, &channelDesc2, w2, h2);
    cudaMemcpyToArray(TargetcuArray, 0, 0, targetgray->imageData,size2,cudaMemcpyHostToDevice);

    texRefTarget.addressMode[0] = cudaAddressModeWrap; //循环寻址方式
    texRefTarget.addressMode[1] = cudaAddressModeWrap;
    texRefTarget.filterMode = cudaFilterModeLinear;

    cudaBindTextureToArray(texRefTarget, TargetcuArray, channelDesc2);

    //----------------------构造从image1----->image2的变换矩阵，存入常量内存c_trans----------//
     float u[9]={-0.00363945, 123.461, 123.154, 1.0105, 0.868716, 1.2, 1.2, -0.000110196, 0.000174003};
     float matM[3][3]={ u[3]*cos(u[0]), u[4]*sin(u[0]),  0,
                    -u[5]*sin(u[0]),  u[6]*cos(u[0]), 0,
                    u[7],             u[8],             1}; 
     CvMat M = cvMat(3,3,CV_32FC1,matM);  //求得的是image2-->image1的M。
     CvMat invM = cvMat(3,3,CV_32FC1,matM);  //求得的是image1-->image2的M。
     cvInvert(&M,&invM);
     for (int i=0;i<3;i++)
     {
         for (int j=0;j<3;j++)
         {
             c_trans[i*3+j]=CV_MAT_ELEM(invM,float,i,j); 
         }
     }
    //-------------------------------------------------------------------------------------------------//

     float *dev_bitmap=NULL;
     int size=size1*sizeof(float);
     cudaMalloc((void**)&dev_bitmap, size);

    //开始变换图像
     dim3 blockSize(ThreadX, ThreadY); 
     dim3 gridSize( (w1+ThreadX-1) / blockSize.x, (h1+ThreadY-1) / blockSize.y);
     transformKernel<<<gridSize, blockSize>>>(dev_bitmap,w1,h1);

     CPUBitmap bitmap(w1,h1);
     cudaMemcpy(bitmap.get_ptr(),(unsigned char*)dev_bitmap,bitmap.image_size(),cudaMemcpyDeviceToHost);

     bitmap.display_and_exit();

     cvWaitKey(0);

     cvReleaseImage(&source);
     cvReleaseImage(&sourcegray);
     cvReleaseImage(&target);
     cvReleaseImage(&targetgray);


     cvDestroyAllWindows();

     cudaFreeArray(SourcecuArray);
     cudaFreeArray(TargetcuArray);
     cudaFree(dev_bitmap);
     return 0;
}