I use cudaMalloc malloc 3X1280X720 PointXYZRGBA.
then I call cuda kernel function.
I can write the first 1280X720 and second 1280X720 memory block.
for the third block I can write and print its value in kernel function, the values aren’t zeros. but I cann’t copy them from gpu to cpu.
I try to copy the third 1280X720 memeory block to cpu and print them, I found all of them are zeros. I don’t know why, could someone know the reason?
the information about my codes are as follow:
cudaMalloc(&Pcd, sizeof(PointXYZRGBA)3bufferDimInput);
struct PointXYZRGBAGL{
float x;
float y;
float z;
float r;
float g;
float b;
};
I use cudaMalloc malloc 31280720 memory space the following function are used :
"cudaMalloc(&D_Point, 3sizeof(PointXYZRGBA)1280720);". After I call a funtion:"extern “C” void GeneratePointCloud_Multi_RGBD1(const uchar4 d_colorInput0, const uint16_t* d_depthinput0, const uchar4* d_colorInput1, const uint16_t* d_depthinput1, float* transformation01, const uchar4* d_colorInput2, const uint16_t* d_depthinput2, float* transformation02, float4 intrinsics, float scale, unsigned int width, unsigned int height, PointXYZRGBAGL D_Point, PointXYZRGBA D_PointCloud);
"
and this function has a kernelfunction:
global void GeneratePointCloud_Multi_RGBD_Kernel1(const uchar4* d_colorInput0, const uint16_t* d_depthinput0, const uchar4* d_colorInput1, const uint16_t* d_depthinput1, float* transformation01, const uchar4* d_colorInput2, const uint16_t* d_depthinput2, float* transformation02, float4 intrinsics, float scale, unsigned int width, unsigned int height, PointXYZRGBAGL D_Point, PointXYZRGBA D_PointCloud)
{
const dim3 gridSize((3*width + T_PER_BLOCK - 1) / T_PER_BLOCK, (height + T_PER_BLOCK - 1) / T_PER_BLOCK);//T_PER_BLOCK = 8
const dim3 blockSize(T_PER_BLOCK, T_PER_BLOCK);
GeneratePointCloud_Multi_RGBD_Kernel1 <<<gridSize, blockSize>>>(d_colorInput0, d_depthinput0, d_colorInput1, d_depthinput1, transformation01, d_colorInput2, d_depthinput2, transformation02, intrinsics, scale, width, height, D_Point, D_PointCloud );
}