Name ARB_copy_image Name Strings GL_ARB_copy_image Contact Michael Gold, NVIDIA Corporation (gold 'at' nvidia.com) Graham Sellers, AMD (graham.sellers 'at' amd.com) Contributors Jeff Bolz, NVIDIA Corporation (jbolz 'at' nvidia.com) James Jones, NVIDIA Corporation (jajones 'at' nvidia.com) Joe Kain, NVIDIA Corporation (jkain 'at' nvidia.com) Benjamin Morris, NVIDIA Corporation (bmorris 'at' nvidia.com) Michael Morrison, NVIDIA Corporation (mmorrison 'at' nvidia.com) Aaron Plattner, NVIDIA Corporation (aplattner 'at' nvidia.com) Mathias Schott, NVIDIA Corporation (mschott 'at' nvidia.com) Thomas Volk, NVIDIA Corporation (tvolk 'at' nvidia.com) Eric Werness, NVIDIA Corporation (ewerness 'at' nvidia.com) Daniel Koch, TransGaming Inc. (daniel 'at' transgaming.com) Notice Copyright (c) 2012-2013 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 2012/06/12. Version Last Modified Date: May 10, 2012 Revision: 3 Number ARB Extension #123 Dependencies OpenGL 1.1 is required. The extension is written against the OpenGL 4.2 (Core Profile) Specification. This extension interacts with OpenGL 4.3 and ARB_texture_view. This extension interacts with the Compatibility Profile. This extension interacts with EXT_texture_compression_s3tc. This extension interacts with OpenGL 3.0 and ARB_texture_compression_rgtc. This extension interacts with OpenGL 4.2 and ARB_texture_compression_bptc. Overview This extension enables efficient image data transfer between image objects (i.e. textures and renderbuffers) without the need to bind the objects or otherwise configure the rendering pipeline. This is accomplised by adding a new entry-point CopyImageSubData, which takes a named source and destination. CopyImageSubData does not perform general-purpose conversions such as scaling, resizing, blending, color-space, or format conversions. It should be considered to operate in a manner similar to a CPU memcpy, but using the GPU for the copy. CopyImageSubData supports copies between images with different internal formats, if the formats are compatible for TextureViews. CopyImageSubData also supports copying between compressed and uncompressed images if the compressed block / uncompressed texel sizes are the same. New Procedures and Functions void CopyImageSubData( uint srcName, enum srcTarget, int srcLevel, int srcX, int srcY, int srcZ, uint dstName, enum dstTarget, int dstLevel, int dstX, int dstY, int dstZ, sizei srcWidth, sizei srcHeight, sizei srcDepth); New Tokens None Additions to Chapter 4 of the OpenGL 4.2 (Core Profile) Specification (Per-Fragment Operations and the Frame Buffer) Append to section 4.3.2 (Copying Pixels): The function void CopyImageSubData( uint srcName, enum srcTarget, int srcLevel, int srcX, int srcY, int srcZ, uint dstName, enum dstTarget, int dstLevel, int dstX, int dstY, int dstZ, sizei srcWidth, sizei srcHeight, sizei srcDepth); may be used to copy a region of texel data between two image objects. An image object may be either a texture or a renderbuffer. CopyImageSubData does not perform general-purpose conversions such as scaling, resizing, blending, color-space, or format conversions. It should be considered to operate in a manner similar to a CPU memcpy. CopyImageSubData can copy between images with different internal formats, provided the formats are compatible. CopyImageSubData also allows copying between certain types of compressed and uncompressed internal formats as detailed in Table 4.X.1. This copy does not perform on-the-fly compression or decompression. When copying from an uncompressed internal format to a compressed internal format, each texel of uncompressed data becomes a single block of compressed data. When copying from a compressed internal format to an uncompressed internal format, a block of compressed data becomes a single texel of uncompressed data. The texel size of the uncompressed format must be the same size the block size of the compressed formats. Thus it is permitted to copy between a 128-bit uncompressed format and a compressed format which uses 8-bit 4x4 blocks, or between a 64-bit uncompressed format and a compressed format which uses 4-bit 4x4 blocks. INVALID_OPERATION is generated if the texel size of the uncompressed image is not equal to the block size of the compressed image. The source object is identified by and . Similarly the destination object is identified by and . The interpretation of the name depends on the value of the corresponding target parameter. If the target parameter is RENDERBUFFER, the name is interpreted as the name of a renderbuffer object. If the target parameter is a texture target, the name is interpreted as a texture object. All non-proxy texture targets are accepted, with the exception of TEXTURE_BUFFER and the cubemap face selectors described in table 3.17. INVALID_ENUM is generated if either target is not RENDERBUFFER or a valid non-proxy texture target, or is TEXTURE_BUFFER, or is one of the cubemap face selectors described in table 3.17, or if the target does not match the type of the object. INVALID_OPERATION is generated if either object is a texture and the texture is not complete (as defined in section 3.9.14), if the source and destination internal formats are not compatible (see below), or if the number of samples do not match. INVALID_VALUE is generated if either name does not correspond to a valid renderbuffer or texture object according to the corresponding target parameter. and identify the source and destination level of detail. For textures, this must be a valid level of detail in the texture object. For renderbuffers, this value must be zero. INVALID_VALUE is generated if the specified level is not a valid level for the image. , , and specify the lower left texel coordinates of a -wide by -high by -deep rectangular subregion of the source texel array. (Compatibility Only: Negative values of , , and correspond to the coordinates of border texels, addressed as in figure 3.10.) Similarly, , and specify the coordinates of a subregion of the destination texel array. The source and destination subregions must be contained entirely within the specified level of the corresponding image objects. The dimensions are always specified in texels, even for compressed texture formats. But it should be noted that if only one of the source and destination textures is compressed then the number of texels touched in the compressed image will be a factor of the block size larger than in the uncompressed image. INVALID_VALUE is generated if the dimensions of the either subregion exceeds the boundaries of the corresponding image object, or if the image format is compressed and the dimensions of the subregion fail to meet the alignment constraints of the format. Slices of a TEXTURE_1D_ARRAY, TEXTURE_2D_ARRAY, TEXTURE_CUBE_MAP_ARRAY TEXTURE_3D and faces of TEXTURE_CUBE_MAP are all compatible provided they share a compatible internal format, and multiple slices or faces may be copied between these objects with a single call by specifying the starting slice with and , and the number of slices to be copied with . Cubemap textures always have six faces which are selected by a zero-based face index, according to the order specified in table 3.17. For the purposes of CopyImageSubData, two internal formats are considered compatible if any of the following conditions are met: * the formats are the same, * the formats are considered compatible according to the compatibility rules used for texture views as defined in section 3.9.X. In particular, if both internal formats are listed in the same entry of Table 3.X.2, they are considered compatible, or * one format is compressed and the other is uncompressed and Table 4.X.1 lists the two formats in the same row. If the formats are not compatible INVALID_OPERATION is generated. ------------------------------------------------------------------------ | Texel / | Uncompressed | | | Block | internal format | Compressed internal format | | size | | | ------------------------------------------------------------------------ | 128-bit | RGBA32UI, | COMPRESSED_RGBA_S3TC_DXT3_EXT, | | | RGBA32I, | COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT,| | | RGBA32F | COMPRESSED_RGBA_S3TC_DXT5_EXT, | | | | COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT,| | | | COMPRESSED_RG_RGTC2, | | | | COMPRESSED_SIGNED_RG_RGTC2, | | | | COMPRESSED_RGBA_BPTC_UNORM, | | | | COMPRESSED_SRGB_ALPHA_BPTC_UNORM, | | | | COMPRESSED_RGB_BPTC_SIGNED_FLOAT, | | | | COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT | ------------------------------------------------------------------------ | 64-bit | RGBA16F, RG32F, | COMPRESSED_RGB_S3TC_DXT1_EXT, | | | RGBA16UI, RG32UI, | COMPRESSED_SRGB_S3TC_DXT1_EXT, | | | RGBA16I, RG32I, | COMPRESSED_RGBA_S3TC_DXT1_EXT, | | | RGBA16, RGBA16_SNORM | COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT,| | | | COMPRESSED_RED_RGTC1, | | | | COMPRESSED_SIGNED_RED_RGTC1 | ------------------------------------------------------------------------ Table 4.X.1: Compatible internal formats for copying between compressed and uncompressed internal formats with CopyImageSubData. Formats in the same row can be copied between each other. Dependencies on OpenGL 4.3 and ARB_texture_view: If OpenGL 4.3 or later and ARB_texture_view are not supported, any references to view-compatible image formats should be removed. Dependencies on the Compatibility Profile: If the Compatibility Profile is not supported, remove any references to negative values for the source dimensions and to border texels. Dependencies on EXT_texture_compression_s3tc: If EXT_texture_compression_s3tc is not supported, remove any references to S3TC compressed texture formats. Dependencies on OpenGL 3.0 and ARB_texture_compression_rgtc: If OpenGL 3.0 or later, and ARB_texture_compression_rgtc is not supported, remove any references to the RGTC compressed texture formats. Dependencies on OpenGL 4.2 and ARB_texture_compression_bptc: If OpenGL 4.2 or later, and ARB_texture_compression_bptc is not supported, remove any references to the BPTC compressed texture formats. Errors CopyImageSubData may fail with any of the following errors: INVALID_ENUM is generated * if either or - is not RENDERBUFFER or a valid non-proxy texture target - is TEXTURE_BUFFER, or - is one of the cubemap face selectors described in table 3.17, * if the target does not match the type of the object. INVALID_OPERATION is generated * if either object is a texture and the texture is not complete, * if the source and destination formats are not compatible, * if the source and destination number of samples do not match, * if one image is compressed and the other is uncompressed and the block size of compressed image is not equal to the texel size of the compressed image. INVALID_VALUE is generated * if either or does not correspond to a valid renderbuffer or texture object according to the corresponding target parameter, or * if the specified level is not a valid level for the image, or * if the dimensions of the either subregion exceeds the boundaries of the corresponding image object, or * if the image format is compressed and the dimensions of the subregion fail to meet the alignment constraints of the format. Sample Code TBD Issues 1) Should there be a single function for all image types, or "per-dimensional" functions? RESOLVED: A single function can support all image types. Not only are per-dimensional functions an unnecessary convenience, they also restrict some of the flexibility offered by this extension, e.g. copying a slice of data between a 2D and a 3D image. 2) Should the extension support "deep copies", i.e. multiple slices of a 3D texture, a 2D array texture, or a cubemap? RESOLVED: Yes, there may be performance advantages in copying a multiple slices in a single call. 3) Should renderbuffers be supported by the same function as textures? RESOLVED: Yes, there is no fundamental difference between the two object classes, and allowing them to be used interchangeably has the advantage of allowing data transfers between them. 4) Is the "target" parameter necessary? RESOLVED: Yes, given the current object model and texture API, there are two obvious applications of the target parameter: 1) Allows the selection of a single cubemap face. 2) Differentiate between TEXTURE and RENDERBUFFER targets. 5) Should the target TEXTURE_CUBE_MAP be supported, and with what behavior? RESOLVED: Given the resolution of issue 7, this is moot. The TEXTURE_CUBE_MAP token is the only way to access a cubemap. 6) Should the "extra" parameters for a given dimension be ignored, or should there be required values? Resolved. All parameters are required to have sensible values, for the simple reason that future extensions may give meaning to these values. For dimensions which are currently superfluous, the offset must be zero and the size must be one, e.g. if the target is TEXTURE_2D, z must be 0 and depth must be 1. 7) Should the per-face cubemap targets be accepted at all? Why not use Z as the selector? Resolved. The existing per-face targets effectively define the face order: face_index = face_target - TEXTURE_CUBE_MAP_POSITIVE_X; Therefore it makes sense to generalize a cubemap as an array of size 6, and use the Z parameter to select the face(s). 8) How is this extension different from NV_copy_image? RESOLVED: This extension does not include the WGL/GLX windowing- layer copy functions. This extension adds the ability to copy between images which have different formats where the formats are compatible for texture views. This extension adds the ability to copy between uncompressed and compressed images if the texel and block sizes are the same. 9) Please explain copying between compressed and uncompressed images. Is the copy doing compression/decompression on the fly? DISCUSSION. The CopyImageSubData does NOT provide any on-the-fly compression, decompression, or even any format conversions. Instead, it provides the ability for an application to implement it's own compressor in a shader by constructing the blocks and writing them out as texels and then copying them to a compressed texture so that it can in turn be used for sampling. (10) Is copying from/to images with ETC2/EAC compressed texture formats defined? RESOLVED: Only when the source and destination formats are identical, because texture view classes for ETC2/EAC are not defined (see issue 13 of the ARB_texture_view specification). Revision History Rev. Date Author Changes ---- -------- -------- ----------------------------------------- 1 04/26/2012 gsellers Initial revision based on NV_copy_image. 2 05/08/2012 dgkoch Update some references for 4.2 spec Allow copies between view-compatible images. Allow copies between compressed/uncompressed images. Clarify errors, updated overview. 3 05/10/2012 dgkoch Clarify dimensions refer to texels in the src image. Re-order/reflow text to make it read better. Add mention of TEXTURE_CUBE_MAP_ARRAY. 4 10/08/2013 Jon Leech Add issue 10 discussing why copying involving conversions between ETC2/EAC compressed formats is not supported (Bug 11011).