extensions/intel/cl_intel_packed_yuv.txt

Name Strings

    cl_intel_packed_yuv

Contributors

    Ben Ashbaugh, Intel

Contact

    Ben Ashbaugh, Intel (ben.ashbaugh 'at' intel.com)

Version

    Version 1, July 14, 2016

Number

    OpenCL Extension #42

Status

    Final Draft

Dependencies

    OpenCL 1.2 is required.  This extension is written against revision
    19 of the OpenCL 1.2 specification.

Overview

    The purpose of this extension is to provide OpenCL support for packed
    YUV images.

New Procedures and Functions

    None

New Tokens

    Accepted as <image_channel_order> of cl_image_format:

       CL_YUYV_INTEL                               0x4076
       CL_UYVY_INTEL                               0x4077
       CL_YVYU_INTEL                               0x4078
       CL_VYUY_INTEL                               0x4079

Additions to Chapter 5 of the OpenCL 1.2 Specification:

    In section 5.3.1.2 "Image Descriptor", add a sentence to the
    end of the description of <image_width>:

   "<image_width> is...  If the image is a packed YUV image, <image_width>
    is the width of the Y data and must be even."

    Add to table 5.6 "List of supported Image Channel Order Values" in
    section 5.3.1.1 "Image Format Descriptor":

   "CL_YUYV_INTEL, CL_UYVY_INTEL, CL_YVYU_INTEL, CL_VYUY_INTEL.  This format
    can only be used if channel data type = CL_UNORM_INT8."

    Add to section 5.3.1.1 "Image Format Descriptor" after the memory layout
    descriptions for CL_RGBA image formats:

   "The only supported <image_channel_data_type> for packed YUV images is
    CL_UNORM_INT8.

    The memory layout of a packed YUV image when <image_channel_order> is
    CL_YUYV_INTEL is:

        +---------+---------+---------+---------+
        |    Y    |    U    |    Y    |    V    |   ...
        | Pixel N | Pixel N |Pixel N+1| Pixel N |
        +---------+---------+---------+---------+
    Byte ->  0         1         2         3
    Offset

    The memory layout of a packed YUV image when <image_channel_order> is
    CL_UYVY_INTEL is:

        +---------+---------+---------+---------+
        |    U    |    Y    |    V    |    Y    |   ...
        | Pixel N | Pixel N | Pixel N |Pixel N+1|
        +---------+---------+---------+---------+
    Byte ->  0         1         2         3
    Offset

    The memory layout of a packed YUV image when <image_channel_order> is
    CL_YVYU_INTEL is:

        +---------+---------+---------+---------+
        |    Y    |    V    |    Y    |    U    |   ...
        | Pixel N | Pixel N |Pixel N+1| Pixel N |
        +---------+---------+---------+---------+
    Byte ->  0         1         2         3
    Offset

    The memory layout of a packed YUV image when <image_channel_order> is
    CL_VYUY_INTEL is:

        +---------+---------+---------+---------+
        |    V    |    Y    |    U    |    Y    |   ...
        | Pixel N | Pixel N | Pixel N |Pixel N+1|
        +---------+---------+---------+---------+
    Byte ->  0         1         2         3
    Offset

    Add a new Table 5.x "Additional Required Image Formats for Read-Only
    2D Images" to section 5.3.2.1 "Minimum List of Supported Image Formats":

   "--------------------------------------------------------------------
    image_num_channels    image_channel_order    image_channel_data_type
    ------------------    -------------------    -----------------------
    2                     CL_YUYV_INTEL          CL_UNORM_INT8

    2                     CL_UYVY_INTEL          CL_UNORM_INT8

    2                     CL_YVYU_INTEL          CL_UNORM_INT8

    2                     CL_VYUY_INTEL          CL_UNORM_INT8
    --------------------------------------------------------------------"

    In section 5.3.3 "Reading, Writing, and Copying Image Objects", add a
    sentence to the end of the descriptions of <origin> and <region> under
    clEnqueueReadImage and clEnqueueWriteImage:

   "<origin> defines... If <image> is a packed YUV image, the x value given
    by <origin>[0] indexes Y data in the image and must be even.

    <region> defines... If <image> is a packed YUV image, the width value
    given by <origin>[0] describes Y data image and must be even."

    Add a sentence to the end of the descriptions of <src_origin>,
    <dst_origin>, and <region> under clEnqueueCopyImage:

   "<src_origin> defines... If <src_image> is a packed YUV image, the x
    value given by <src_origin>[0] indexes Y data in <src_image> and must
    be even.

    <dst_origin> defines... If <dst_image> is a 2D image object, the z
    value given by <dst_origin>[2] must be 0.  If <dst_image> is a packed
    YUV image, the x value given by <dst_origin>[0] indexes Y data in
    <dst_image> and must be even.

    <region> defines... If <src_image> or <dst_image> is a packed YUV
    image, the <width> value given by <region>[0] describes Y data and
    must be even."

    In section 5.3.5 "Copying between Image and Buffer Objects", add a
    sentence to the end of the descriptions of <src_origin> and <region>
    for clEnqueueCopyImageToBuffer:

   "<src_origin> defines... If <src_image> is a packed YUV image, the x
    value given by <origin>[0] indexes Y data in the image and must be
    even.

    <region> defines... If <src_image> is a packed YUV image, the width
    value given by <origin>[0] describes Y data and must be even."

    Add a sentence to the end of the descriptions of <dst_origin> and
    <region> for clEnqueueCopyBufferToImage:

   "<dst_origin> defines... If <dst_image> is a packed YUV image, the x
    value given by <origin>[0] indexes Y data in the image and must be
    even.

    <region> defines... If <dst_image> is a packed YUV image, the width
    value given by <region>[0] describes Y data and must be even."

    In section 5.3.6 "Mapping Image Objects", add a sentence to the end
    of the descriptions of <origin> and <region> for clEnqueueMapImage:

   "<origin> defines... If <image> is a packed YUV image, the x value
    given by <origin>[0] describes Y data and must be even.

    <region> defines... If <image> is a packed YUV image, the width
    value given by <region>[0] describes Y data and must be even."

Additions to Chapter 6 of the OpenCL 1.2 Specification:

    In section 6.12.14.1.1 "Determining the border color", add the
    packed YUV formats to the second bullet describing the cases where
    the border color is (0.0f, 0.0f, 0.0f, 1.0f):

   "  * If image channel order is CL_R, CL_RG, CL_RGB, CL_LUMINANCE,
        CL_YUYV_INTEL, CL_UYVY_INTEL, CL_YVYU_INTEL, or CL_VYUY_INTEL,
        the border color is (0.0f, 0.0f, 0.0f, 1.0f)."

    Add to the un-numbered table in section 6.12.14.6 "Mapping image
    channels to color values returned by read_image and color values
    passed to write_image to image channels":

   "--------------------------------------------------------------
                                     float4, int4 or unsigned int4
    Channel Order                       components of channel data
    -----------------------------    -----------------------------
    CL_YUYV_INTEL, CL_UYVY_INTEL,                 ( V, Y, U, 1.0 )
    CL_YVYU_INTEL, CL_VYUY_INTEL
    --------------------------------------------------------------"

Revision History

    Version 1 - Initial Revision