Name

     ARB_get_texture_sub_image

Name Strings

     GL_ARB_get_texture_sub_image

Contact

     Brian Paul, VMware Inc.  (brianp 'at' vmware.com)

Contributors

     Brian Paul
     Daniel Koch, NVIDIA

Notice

    Copyright (c) 2014 The Khronos Group Inc. Copyright terms at
        http://www.khronos.org/registry/speccopyright.html

Specification Update Policy

    Khronos-approved extension specifications are updated in response to
    issues and bugs prioritized by the Khronos OpenGL Working Group. For
    extensions which have been promoted to a core Specification, fixes will
    first appear in the latest version of that core Specification, and will
    eventually be backported to the extension document. This policy is
    described in more detail at
        https://www.khronos.org/registry/OpenGL/docs/update_policy.php

Status

    Complete. 
    Approved by the ARB on June 26, 2014.
    Ratified by the Khronos Board of Promoters on August 7, 2014.

Version

    Date: June 16, 2014
    Revision: 7

Number

    ARB Extension #165

Dependencies

    OpenGL 2.0 is required

    The extension is written against the OpenGL 4.4 Specification, Core
    Profile, March 19, 2014.

Overview

    This extension adds a new function to get sub-regions of texture
    images.

IP Status

    No known IP claims.

New Procedures and Functions

    void GetTextureSubImage(uint texture, int level,
                            int xoffset, int yoffset, int zoffset,
                            sizei width, sizei height, sizei depth,
                            enum format, enum type,
                            sizei bufSize, void *pixels);

    void GetCompressedTextureSubImage(uint texture,
                                      int level, int xoffset,
                                      int yoffset, int zoffset,
                                      sizei width, sizei height,
                                      sizei depth, sizei bufSize,
                                      void *pixels);

New Types

    None

New Tokens

    None

Additions to Chapter 8 of the OpenGL 4.4 (Core Profile) Specification (Textures and Samplers)

    Modify section 8.11.4, Texture Image Queries p. 224

    (insert before the discussion of GetCompressedTexImage)

    Sub-regions of a texture image may be obtained from a texture object
    with the command

        void GetTextureSubImage(uint texture, int level,
                               int xoffset, int yoffset, int zoffset,
                               sizei width, sizei height, sizei depth,
                               enum format, enum type,
                               sizei bufSize, void *pixels);

    <texture> is the name of the source texture object. and must not be a
    buffer or multisample texture. The effective <target> parameter is the
    value of TEXTURE_TARGET for texture. <level>, <format>, <type> and
    <pixels> have the same meaning as for GetTexImage. <bufSize> is the size
    of the buffer to receive the retrieved pixel data.

    For cube map textures, the behavior is as though GetTextureImage were
    called, but only texels from the requested cube map faces (selected by
    <zoffset> and <depth>, as described below) were returned.

    <xoffset>, <yoffset> and <zoffset> values indicate the position of the
    subregion to return. <width>, <height> and <depth> indicate the size of
    the region to return. These parameters have the same meaning as for
    TexSubImage3D, though for one- and two-dimensional textures there are
    extra restrictions, described in the errors section below).

    For one-dimensional array textures, <yoffset> is interpreted as the
    first layer to access and <height> is the number of layers to access.
    For two-dimensional array textures, <zoffset> is interpreted as the
    first layer to access and <depth> is the number of layers to access.
    Cube map textures are treated as an array of six slices in the
    z-dimension, where the value of <zoffset> is interpreted as specifying
    the cube map face for the corresponding layer in table 9.2 and <depth>
    is the number of faces to access. For cube map array textures, <zoffset>
    is the first layer-face to access, and <depth> is the number of layer-
    faces to access. Each layer-face is translated into an array layer and a
    cube map face as described for layer-face numbers in section 8.5.3.

    Component groups from the specified sub-region are packed and placed
    into memory as described for GetTextureImage, starting with the texel at
    (<xoffset>, <yoffset>, <zoffset>).

    Errors

    An INVALID_OPERATION error is generated if <texture> is not the name of
    an existing texture object.

    An INVALID_OPERATION error is generated if <texture> is the name of a
    buffer or multisample texture.

    An INVALID_VALUE error is generated if <xoffset>, <yoffset> or <zoffset>
    are negative.

    An INVALID_VALUE error is generated by if <xoffset> + <width> is greater
    than the texture level's width or if <yoffset> + <height> is greater
    than the texture level's height or if <zoffset> + <depth> is greater
    than the texture level's depth.

    An INVALID_VALUE error is generated if the effective <target> is
    TEXTURE_1D and either <yoffset> is not zero, or <height> is not one.

    An INVALID_VALUE error is generated if the effective <target> is
    TEXTURE_1D, TEXTURE_1D_ARRAY, TEXTURE_2D, or TEXTURE_RECTANGLE and
    either <zoffset> is not zero, or <depth> is not one.

    An INVALID_OPERATION error is generated if the buffer size required to
    store the requested data is greater than <bufSize>.


    (insert after the discussion of GetCompressedTexSubImage and before
    section 8.12 begins)

    If the compressed data are arranged into fixed-size blocks of texels,
    the command

        void GetCompressedTextureSubImage(uint texture,
                                          int level, int xoffset,
                                          int yoffset, int zoffset,
                                          sizei width, sizei height,
                                          sizei depth, sizei bufSize,
                                          void *pixels);

    can be used to obtain a sub-region of a compressed texture image instead
    of the whole image. <texture> is the name of the source texture object,
    and must not be a buffer or multisample texture. The effective <target>
    is the value of TEXTURE_TARGET for <texture>. <level> and <pixels> have
    the same meaning as the corresponding arguments CompressedTexSubImage3D.
    <bufSize> indicates the size of the buffer to receive the retrieved
    pixel data.

    For cube map textures, the behavior is as though GetCompressedTexImage
    were called once for each requested face (selected by <zoffset> and
    <depth>, as described below) with <target> corresponding to the
    requested texture cube map face as indicated by Table 9.2. <pixels> is
    offset appropriately for each successive image.

    <xoffset>, <yoffset> and <zoffset> indicate the position of the
    subregion to return. <width>, <height> and <depth> indicate the size of
    the region to return. These arguments have the same meaning as for
    CompressedTexSubImage3D, though there are extra restrictions, described
    in the errors section below.

    The mapping between the <xoffset>, <yoffset>, <zoffset>, <width>,
    <height>, and <depth> parameters and the faces, layers, and layer-faces
    for cube map, array, and cube map array textures is the same as for
    GetTextureSubImage.

    The <xoffset>, <yoffset>, <zoffset>, <width>, <height> and <depth>
    values and must be multiples of the values of
    PACK_COMPRESSED_BLOCK_WIDTH, PACK_COMPRESSED_BLOCK_HEIGHT, and
    PACK_COMPRESSED_BLOCK_DEPTH respectively, unless the offset is zero and
    the corresponding size is the same as the texture size in that
    dimension.

    Pixel storage modes are treated as for GetCompressedTexSubImage. The
    texel at (<xoffset>, <yoffset>, <zoffset>) will be stored at the
    location indicated by <pixels> and the current pixel packing parameters.

    Errors

    In addition to the same errors generated by GetTextureSubImage with
    corresponding parameters:

    An INVALID_VALUE error is generated if <xoffset>, <yoffset> or <zoffset>
    is not a multiple of the compressed block width, height or depth.

    An INVALID_VALUE error is generated if <width>, <height> or <depth> is
    not a multiple of the compressed block width, height or depth
    respectively, unless the offset is zero and the size equals the texture
    image size.

Additions to the AGL/EGL/GLX/WGL Specifications

    None

Dependencies on GL and ES profiles, versions, and other extensions

    If texture arrays are not supported, ignore references to
    1D and 2D-array textures

    If cube map textures are not supported, ignore references to
    cube map textures.

    If cube map arrays texture are not supported, ignore references to
    cube map array textures.

    If texture rectangles are not supported, ignore references to
    rectangle textures.

    If texture compression is not supported, ignore references to
    GetCompressedTexSubImage.

Dependencies on ARB_direct_state_access

    If ARB_direct_state_access is not supported, the TEXTURE_TARGET
    is the <target> that <texture> was initially bound with.

New State

    None.

New Implementation Dependent State

    None.

Conformance Testing

    1) For each type of texture, create a test texture.  Read back the
    whole texture into a buffer with glGet[Compressed]TexImage.  Then
    read back the whole texture into a second buffer, but in sub-image
    pieces (ex: 4x4 grid).  Then compare the two result buffers - they
    should be identical.

Issues

    1) Why is this extension necessary?

    RESOLVED:  In some cases one can wrap a texture with an FBO and
    read back sub-regions with glReadPixels.  However, that doesn't
    work for all texture formats (esp. compressed textures).

    For some textures such as 2D array textures there's no way to obtain
    just one slice of a texture.  It's very inefficient to get the whole
    texture when just one slice is needed.

    An alternate approach to the problem of reading back sub textures
    is with an extension such as GL_INTEL_map_texture.  However, mapping
    textures involves several difficult issues.  Extending glGetTexImage
    to handle sub-regions is much simpler.

    2) What is the behavior of these commands if <bufSize> is exceeded
    when a pixel pack buffer object is bound?

    DISCUSSION: When client memory is being used, INVALID_OPERATION is
    generated if writing more than <bufSize> bytes is necessary to return
    all the requested data. Similarly, when a PBO is bound,
    INVALID_OPERATION is generated if more than <bufSize> bytes are
    required. Note that it can also be an error if *less* than <bufSize>
    bytes would be required when using a PBO if the space between the
    offset indicated by <pixels> and the end of the buffer is insufficient
    for the requested data. This behavior is inherited from GetTexImage
    and GetCompressedTexImage.

    3) Is the <target> parameter needed?

    RESOLVED. No. The target is determined by the texture type.
    For cube map textures the face target is determined by the
    <zoffset> and <depth> parameters by mapping layer numbers to
    cube map texture faces as per Table 9.2. This is consisten
    with the DSA variants of GetTextureImage in ARB_direct_state_access
    which do not include it.

Revision History

    Revision 1, 2014/02/03
       - Initial revision

    Revision 2, 2014/04/10
       - Converted GetTexSubImage and GetTexCompressedSubImage to
         DSA (Direct State Access) style.
       - Added bufSize parameter, per GL_ARB_robustness convention.

    Revision 3, 2014/05/26 - dkoch
       - convert to ARB extension
       - make function naming and parameter lists consistent
       - describe the <bufSize> parameter (issue 2).
       - remove the <target> parameter (issue 3) and fallout.
       - better describe valid texture types and parameter meanings
         for the various texture types.

    Revision 4, 2014/05/30 - dkoch
       - describe implied <target> parameter for cube map textures
       - fix remove incorrect <bufSize> language (Bug 12321)
       - Add additional error language
       - improve issues 2 and 3

    Revision 5, 2014/06/03 - Jon Leech
       - Use 'pixels' instead of 'img' as formal parameter name for
         consistency with API spec / dsa.

    Revision 6, 2014/06/12 - Jon Leech
       - Sync with core API spec language and errors.

    Revision 6, 2014/06/16 - Jon Leech
       - Require GetTextureSubImage for cube map textures to respect the
         three-dimensional pixel pack state, by referring to GetTextureImage
         language (Bug 12329).