Understanding cudaChannelDesc through Driver API what is x, y, z, w?

Dearl all.

I came acrosss the connections betewen cudaMallocArray and cuArrayCreate.

As arguments, there are

struct cudaChannelFormatDesc {
int x, y, z, w;
enum cudaChannelFormatKind f;
};

in cudaMallocArray, and

typedef struct
{
    // dimensions
        unsigned int Width;
        unsigned int Height;
    // format
        CUarray_format Format;
     // channels per array element
        unsigned int NumChannels;
} CUDA_ARRAY_DESCRIPTOR;, with 

// array formats
//
typedef enum CUarray_format_enum {
CU_AD_FORMAT_UNSIGNED_INT8 = 0x01,
CU_AD_FORMAT_UNSIGNED_INT16 = 0x02,
CU_AD_FORMAT_UNSIGNED_INT32 = 0x03,
CU_AD_FORMAT_SIGNED_INT8 = 0x08,
CU_AD_FORMAT_SIGNED_INT16 = 0x09,
CU_AD_FORMAT_SIGNED_INT32 = 0x0a,
CU_AD_FORMAT_HALF = 0x10,
CU_AD_FORMAT_FLOAT = 0x20
} CUarray_format;

First I would like make sure the CUDA_ARRAY_DESCRIPTOR is corresponding to cudaChannelFormatDesc in runtime API.

Thus for given Runtime API:

cudaMallocArray ( array, desc, width, height),

the following drvier API do the same thing.

struct CUDA_ARRAY_DESCRIPTOR cu_desc;
and
cuArrayCreate(pHandle, cu_desc).

At this time how the field x, y, z, w of cudaChannelFormatDesc are used in making cu_desc?

And I would like to ask the following correspondences are right.
Runtime API --------- Driver API
cudaChannelFormatKindSigned ------ CU_AD_FORMAT_SIGNED_XXX
cudaChannelFormatKindUnsigned ------ CU_AD_FORMAT_UNSIGNED_XXX
cudaChannelFormatKindFloat ------ CU_AD_FORMAT_FLOAT

Thanks for reading and replying in advance.

S.

Found some information from below:

[url=“http://theinf2.informatik.uni-jena.de/theinf2_multimedia/Lectures_material/CUDA_WS08_09/CUDA_Lecture8_181108-view_image-1-called_by-theinf2-original_site--original_page-41-p-101.ppt”]http://theinf2.informatik.uni-jena.de/thei...ge-41-p-101.ppt[/url]

It says :
x,y,z,w: #bits per component

Slide 11 shows:
CUDA Driver API
Declare: CUarray
Channel: CUDA_ARRAY_DESCRIPTOR object
Allocate: cuArrayCreate
Copy (from linear): CUDA_MEMCPY2D object
Free: cuArrayDestroy
Which caused a question how to use CUDA_MEMCPY2D in cuMemcpyHtoA in 2D case.

Hope having lots of helpers for driver API too.

S.