cudaEventCreate（）的输入：参数太少

Question

我在vs2015中使用cuda 9.0。以下是一些导致错误Too few arguments的cuda代码：

cudaEvent_t start, end;
cudaEventCreate(&start);
cudaEventCreate(&end);

这些代码位于函数calcLBPGPU（https://github.com/nourani/LBP/blob/master/LBPGPU.cu）中，如下所示：

void calcLBPGPU( const unsigned char * h_src, unsigned char * h_dst, const int width, const int height,
            const LBPMapping * mapping ) {

    unsigned char *d_Src = NULL, *d_Dst = NULL;

    cudaMalloc( &d_Src, sizeof(char) * height * width );
    cudaMalloc( &d_Dst, sizeof(char) * height * width );
    cudaMemset( (void *) d_Dst, 0, sizeof(char) * height * width );
    cudaMemcpy( d_Src, h_src, sizeof(char) * height * width, cudaMemcpyHostToDevice );

    dim3 numThreadsPerBlock, numBlocks;
    numThreadsPerBlock.x = width;
//  numThreadsPerBlock.y    = height;
    numBlocks.x = height;
//  numBlocks.y             = 1;

    cudaEvent_t start, end;
    cudaEventCreate(&start);
    cudaEventCreate(&end);
    float time;

    cout << "before gpu call" << endl;
    cudaEventRecord( start, 0 );
    calcLBPKernel<<< numBlocks, numThreadsPerBlock >>>( d_Src, d_Dst, width, height, mapping );
    cudaEventRecord( end, 0 );
    cudaEventSynchronize( end );
    cudaEventElapsedTime( &time, start, end );
    cudaEventDestroy( start );
    cudaEventDestroy( end );
    cout << "after gpu sync. Took " << time / 1000 << "s" << endl;

    cudaMemcpy( h_dst, d_Dst, sizeof(char) * height * width, cudaMemcpyDeviceToHost );
    cudaFree( d_Src );
    cudaFree( d_Dst );

}