Name

    EXT_texture_compression_latc

Name Strings

    GL_EXT_texture_compression_latc
    GL_NV_texture_compression_latc (legacy)

Contributors

    Mark J. Kilgard, NVIDIA
    Pat Brown, NVIDIA
    Yanjun Zhang, S3
    Attila Barsi, Holografika
    Daniel Koch, NVIDIA
     
Contact

    Mark J. Kilgard, NVIDIA Corporation (mjk 'at' nvidia.com)

Status

    Shipping for GeForce 8 Series (November 2006)

Version

    Last Modified Date:         July 30, 2017
    Revision: 2

Number

    331

Dependencies

    OpenGL 1.3 or ARB_texture_compression required

    This extension is written against the OpenGL 2.0 (September 7,
    2004) specification.

    This extension interacts with OpenGL 2.0 and ARB_texture_non_power_of_two.

Overview

    This extension introduces four new block-based texture compression
    formats suited for unsigned and signed luminance and luminance-alpha
    textures (hence the name "latc" for Luminance-Alpha Texture
    Compression).

    These formats are designed to reduce the storage requirements and
    memory bandwidth required for luminance and luminance-alpha textures
    by a factor of 2-to-1 over conventional uncompressed luminance and
    luminance-alpha textures with 8-bit components (GL_LUMINANCE8 and
    GL_LUMINANCE8_ALPHA8).

    The compressed signed luminance-alpha format is reasonably suited
    for storing compressed normal maps.

New Procedures and Functions

    None.

New Tokens

    Accepted by the <internalformat> parameter of TexImage2D,
    CopyTexImage2D, and CompressedTexImage2D and the <format> parameter
    of CompressedTexSubImage2D:

        COMPRESSED_LUMINANCE_LATC1_EXT                 0x8C70
        COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT          0x8C71
        COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT           0x8C72
        COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT    0x8C73

Additions to Chapter 2 of the OpenGL 2.0 Specification (OpenGL Operation)

    None.

Additions to Chapter 3 of the OpenGL 2.0 Specification (Rasterization)

 -- Section 3.8.1, Texture Image Specification

    Add to Table 3.17 (page 155):  Specific compressed internal formats

        Compressed Internal Format                   Base Internal Format
        -------------------------------------------  --------------------
        COMPRESSED_LUMINANCE_LATC1_EXT               LUMINANCE
        COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT        LUMINANCE
        COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT         LUMINANCE_ALPHA
        COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT  LUMINANCE_ALPHA

 -- Section 3.8.2, Alternative Texture Image Specification Commands

    Add to the end of the section (page 163):

    "If the internal format of the texture image being modified is
    COMPRESSED_LUMINANCE_LATC1_EXT, COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT,
    COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT, or
    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT, the texture is stored 
    using one of the two LATC compressed texture image encodings (see
    appendix).  Such images are easily edited along 4x4 texel boundaries,
    so the limitations on TexSubImage2D or CopyTexSubImage2D parameters
    are relaxed.  TexSubImage2D and CopyTexSubImage2D will result in
    an INVALID_OPERATION error only if one of the following conditions
    occurs:

        * <width> is not a multiple of four, <width> plus <xoffset> is not
          equal to TEXTURE_WIDTH, and either <xoffset> or <yoffset> is
          non-zero;

        * <height> is not a multiple of four, <height> plus <yoffset> is not
          equal to TEXTURE_HEIGHT, and either <xoffset> or <yoffset> is
          non-zero; or

        * <xoffset> or <yoffset> is not a multiple of four.

    The contents of any 4x4 block of texels of an LATC compressed texture
    image that does not intersect the area being modified are preserved
    during valid TexSubImage2D and CopyTexSubImage2D calls."

 -- Section 3.8.3, Compressed Texture Images

    Add after the 4th paragraph (page 164) at the end of the
    CompressedTexImage discussion:

    "If <internalformat> is COMPRESSED_LUMINANCE_LATC1_EXT,
    COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT,
    COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT, or
    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT, the compressed texture is 
    stored using one of several LATC compressed texture image formats.
    The LATC texture compression algorithm supports only 2D images
    without borders.  CompressedTexImage1D and CompressedTexImage3D
    produce an INVALID_ENUM error if <internalformat> is an LATC format.
    CompressedTexImage2D will produce an INVALID_OPERATION error if
    <border> is non-zero.

    Add to the end of the section (page 166) at the end of the
    CompressedTexSubImage discussion:

    "If the internal format of the texture image being modified is
    COMPRESSED_LUMINANCE_LATC1_EXT, COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT,
    COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT, or
    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT, the texture is stored
    using one of the several LATC compressed texture image formats.
    Since the LATC texture compression algorithm supports only 2D images,
    CompressedTexSubImage1D and CompressedTexSubImage3D produce an
    INVALID_ENUM error if <format> is an LATC format.  Since LATC images
    are easily edited along 4x4 texel boundaries, the limitations on
    CompressedTexSubImage2D are relaxed.  CompressedTexSubImage2D will
    result in an INVALID_OPERATION error only if one of the following
    conditions occurs:

        * <width> is not a multiple of four, and <width> plus <xoffset> is not
          equal to TEXTURE_WIDTH;

        * <height> is not a multiple of four, and <height> plus <yoffset> is
          not equal to TEXTURE_HEIGHT; or

        * <xoffset> or <yoffset> is not a multiple of four.

    The contents of any 4x4 block of texels of an LATC compressed texture
    image that does not intersect the area being modified are preserved
    during valid TexSubImage2D and CopyTexSubImage2D calls."

Additions to Chapter 4 of the OpenGL 2.0 Specification (Per-Fragment
Operations and the Frame Buffer)

    None.

Additions to Chapter 5 of the OpenGL 2.0 Specification (Special Functions)

    None.

Additions to Chapter 6 of the OpenGL 2.0 Specification (State and
State Requests)

    None.

Additions to Appendix A of the OpenGL 2.0 Specification (Invariance)

    None.

Additions to the AGL/GLX/WGL Specifications

    None.

GLX Protocol

    None.

Dependencies on ARB_texture_compression

    If ARB_texture_compression is supported, all the
    errors and accepted tokens for CompressedTexImage1D,
    CompressedTexImage2D, CompressedTexImage3D, CompressedTexSubImage1D,
    CompressedTexSubImage2D, and CompressedTexSubImage3D also apply
    respectively to the ARB-suffixed CompressedTexImage1DARB,
    CompressedTexImage2DARB, CompressedTexImage3DARB,
    CompressedTexSubImage1DARB, CompressedTexSubImage2DARB, and
    CompressedTexSubImage3DARB.

Dependencies on OpenGL 2.0 or ARB_texture_non_power_of_two

    If OpenGL 2.0 or ARB_texture_non_power_of_two is supported, compressed
    texture images can have sizes that are neither multiples of four nor small
    values like one or two.  The original version of this specification didn't
    allow TexSubImage2D and CompressedTexSubImage2D to update only a portion
    of such images.  The spec has been updated to allow such edits in the
    spirit of the resolution of issue (3) of the EXT_texture_compression_s3tc
    specification.  See the "Implementation Note" section for more details.

Errors

    INVALID_ENUM is generated by CompressedTexImage1D
    or CompressedTexImage3D if <internalformat> is
    COMPRESSED_LUMINANCE_LATC1_EXT, COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT,
    COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT, or
    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT.

    INVALID_OPERATION is generated by CompressedTexImage2D
    if <internalformat> is COMPRESSED_LUMINANCE_LATC1_EXT,
    COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT,
    COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT, or
    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT and <border> is not
    equal to zero.

    INVALID_ENUM is generated by CompressedTexSubImage1D
    or CompressedTexSubImage3D if <format> is
    COMPRESSED_LUMINANCE_LATC1_EXT, COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT,
    COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT, or
    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT.

    INVALID_OPERATION is generated by TexSubImage2D or CopyTexSubImage2D if
    TEXTURE_INTERNAL_FORMAT is COMPRESSED_LUMINANCE_LATC1_EXT,
    COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT,
    COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT, or
    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT and any of the following
    apply:

        * <width> is not a multiple of four, <width> plus <xoffset> is not
          equal to TEXTURE_WIDTH, and either <xoffset> or <yoffset> is
          non-zero;

        * <height> is not a multiple of four, <height> plus <yoffset> is not
          equal to TEXTURE_HEIGHT, and either <xoffset> or <yoffset> is
          non-zero; or

        * <xoffset> or <yoffset> is not a multiple of four.

    INVALID_OPERATION is generated by CompressedTexSubImage2D if
    TEXTURE_INTERNAL_FORMAT is COMPRESSED_LUMINANCE_LATC1_EXT,
    COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT,
    COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT, or
    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT and any of the following
    apply:

        * <width> is not a multiple of four, and <width> plus <xoffset> is not
          equal to TEXTURE_WIDTH;

        * <height> is not a multiple of four, and <height> plus <yoffset> is
          not equal to TEXTURE_HEIGHT; or

        * <xoffset> or <yoffset> is not a multiple of four.


    The following restrictions from the ARB_texture_compression
    specification do not apply to LATC texture formats, since subimage
    modification is straightforward as long as the subimage is properly
    aligned.

    DELETE: INVALID_OPERATION is generated by TexSubImage1D, TexSubImage2D,
    DELETE: TexSubImage3D, CopyTexSubImage1D, CopyTexSubImage2D, or
    DELETE: CopyTexSubImage3D if the internal format of the texture image is
    DELETE: compressed and <xoffset>, <yoffset>, or <zoffset> does not equal
    DELETE: -b, where b is value of TEXTURE_BORDER.

    DELETE: INVALID_VALUE is generated by CompressedTexSubImage1D,
    DELETE: CompressedTexSubImage2D, or CompressedTexSubImage3D if the
    DELETE: entire texture image is not being edited:  if <xoffset>,
    DELETE: <yoffset>, or <zoffset> is greater than -b, <xoffset> + <width> is
    DELETE: less than w+b, <yoffset> + <height> is less than h+b, or <zoffset>
    DELETE: + <depth> is less than d+b, where b is the value of
    DELETE: TEXTURE_BORDER, w is the value of TEXTURE_WIDTH, h is the value of
    DELETE: TEXTURE_HEIGHT, and d is the value of TEXTURE_DEPTH.

    See also errors in the GL_ARB_texture_compression specification.

New State

    4 new state values are added for the per-texture object
    GL_TEXTURE_INTERNAL_FORMAT state.

    In the "Textures" state table( page 278), increment the
    TEXTURE_INTERNAL_FORMAT subscript for Z by 4 in the "Type" row.

    [NOTE: The OpenGL 2.0 specification actually should read "n x Z48*"
    because of the 6 generic compressed internal formats in table 3.18.]

New Implementation Dependent State

    None

Appendix

    LATC Compressed Texture Image Formats

    Compressed texture images stored using the LATC compressed image
    encodings are represented as a collection of 4x4 texel blocks,
    where each block contains 64 or 128 bits of texel data.  The image
    is encoded as a normal 2D raster image in which each 4x4 block is
    treated as a single pixel.  If an LATC image has a width or height
    that is not a multiple of four, the data corresponding to texels 
    outside the image are irrelevant and undefined.

    When an LATC image with a width of <w>, height of <h>, and block
    size of <blocksize> (8 or 16 bytes) is decoded, the corresponding
    image size (in bytes) is:
    
        ceil(<w>/4) * ceil(<h>/4) * blocksize.

    When decoding an LATC image, the block containing the texel at offset
    (<x>, <y>) begins at an offset (in bytes) relative to the base of the
    image of:

        blocksize * (ceil(<w>/4) * floor(<y>/4) + floor(<x>/4)).

    The data corresponding to a specific texel (<x>, <y>) are extracted
    from a 4x4 texel block using a relative (x,y) value of
    
        (<x> modulo 4, <y> modulo 4).

    There are four distinct LATC image formats:


    COMPRESSED_LUMINANCE_LATC1:  Each 4x4 block of texels consists of
    64 bits of unsigned luminance image data.  

    Each luminance image data block is encoded as a sequence of 8 bytes,
    called (in order of increasing address):

            lum0, lum1, bits_0, bits_1, bits_2, bits_3, bits_4, bits_5

        The 6 "bits_*" bytes of the block are decoded into a 48-bit bit
        vector:

            bits   = bits_0 +
                     256 * (bits_1 +
                            256 * (bits_2 +
                                   256 * (bits_3 +
                                          256 * (bits_4 + 
                                                 256 * bits_5))))
        
        lum0 and lum1 are 8-bit unsigned integers that are unpacked to
        luminance values LUM0 and LUM1 as though they were pixels with
        a <format> of LUMINANCE and a type of UNSIGNED_BTYE.

        bits is a 48-bit unsigned integer, from which a three-bit control
        code is extracted for a texel at location (x,y) in the block
        using:

            code(x,y) = bits[3*(4*y+x)+2..3*(4*y+x)+0]
        
        where bit 47 is the most significant and bit 0 is the least
        significant bit.

        The luminance value L for a texel at location (x,y) in the block
        is given by:

            LUM0,              if lum0 > lum1 and code(x,y) == 0
            LUM1,              if lum0 > lum1 and code(x,y) == 1
            (6*LUM0+  LUM1)/7, if lum0 > lum1 and code(x,y) == 2
            (5*LUM0+2*LUM1)/7, if lum0 > lum1 and code(x,y) == 3
            (4*LUM0+3*LUM1)/7, if lum0 > lum1 and code(x,y) == 4
            (3*LUM0+4*LUM1)/7, if lum0 > lum1 and code(x,y) == 5
            (2*LUM0+5*LUM1)/7, if lum0 > lum1 and code(x,y) == 6
            (  LUM0+6*LUM1)/7, if lum0 > lum1 and code(x,y) == 7

            LUM0,              if lum0 <= lum1 and code(x,y) == 0
            LUM1,              if lum0 <= lum1 and code(x,y) == 1
            (4*LUM0+  LUM1)/5, if lum0 <= lum1 and code(x,y) == 2
            (3*LUM0+2*LUM1)/5, if lum0 <= lum1 and code(x,y) == 3
            (2*LUM0+3*LUM1)/5, if lum0 <= lum1 and code(x,y) == 4
            (  LUM0+4*LUM1)/5, if lum0 <= lum1 and code(x,y) == 5
            MINLUM,            if lum0 <= lum1 and code(x,y) == 6
            MAXLUM,            if lum0 <= lum1 and code(x,y) == 7

        MINLUM and MAXLUM are 0.0 and 1.0 respectively.

    Since the decoded texel has a luminance format, the resulting RGBA
    value for the texel is (L,L,L,1).


    COMPRESSED_SIGNED_LUMINANCE_LATC1:  Each 4x4 block of texels consists
    of 64 bits of signed luminance image data.  The luminance values of
    a texel are extracted in the same way as COMPRESSED_LUMINANCE_LATC1
    except lum0, lum1, LUM0, LUM1, MINLUM, and MAXLUM are signed values
    defined as follows:

        lum0 and lum1 are 8-bit signed (two's complement) integers.

               { lum0 / 127.0, lum0 > -128
        LUM0 = {
               { -1.0,         lum0 == -128

               { lum1 / 127.0, lum1 > -128
        LUM1 = {
               { -1.0,         lum1 == -128

        MINLUM = -1.0

        MAXLUM =  1.0

    CAVEAT for signed lum0 and lum1 values: the expressions "lum0 >
    lum1" and "lum0 <= lum1" above are considered undefined (read: may
    vary by implementation) when lum0 equals -127 and lum1 equals -128,
    This is because if lum0 were remapped to -127 prior to the comparison
    to reduce the latency of a hardware decompressor, the expressions
    would reverse their logic.  Encoders for the signed LA formats should
    avoid encoding blocks where lum0 equals -127 and lum1 equals -128.


    COMPRESSED_LUMINANCE_ALPHA_LATC2:  Each 4x4 block of texels consists
    of 64 bits of compressed unsigned luminance image data followed by
    64 bits of compressed unsigned alpha image data.

    The first 64 bits of compressed luminance are decoded exactly like
    COMPRESSED_LUMINANCE_LATC1 above.

    The second 64 bits of compressed alpha are decoded exactly like
    COMPRESSED_LUMINANCE_LATC1 above except the decoded value L for this
    second block is considered the resulting alpha value A.

    Since the decoded texel has a luminance-alpha format, the resulting
    RGBA value for the texel is (L,L,L,A).


    COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2:  Each 4x4 block of texels
    consists of 64 bits of compressed signed luminance image data followed
    by 64 bits of compressed signed alpha image data.

    The first 64 bits of compressed luminance are decoded exactly like
    COMPRESSED_SIGNED_LUMINANCE_LATC1 above.

    The second 64 bits of compressed alpha are decoded exactly like
    COMPRESSED_SIGNED_LUMINANCE_LATC1 above except the decoded value L
    for this second block is considered the resulting alpha value A.

    Since this image has a luminance-alpha format, the resulting RGBA
    value is (L,L,L,A).

Issues

    1)  What should these new formats be called?

        RESOLVED: "latc" for Luminance-Alpha Texture Compression.

    2)  How should the uncompressed and filtered texels be returned by
        texture fetches?

        RESOLVED:  Luminance values show up as they do conventionally as
        (L,L,L,1) where the luminance value L is replicated in the red,
        green, and blue components and alpha is forced to 1.  Likewise,
        luminance-alpha values show up as (L,L,L,A) where A is the alpha
        value.

        Alternatively, prior extensions such as NV_float_buffer and
        NV_texture_shader have introduced formats such as GL_FLOAT_R_NV
        and GL_DSDT_NV where one- and two-component texture formats show
        up as (X,0,0,1) or (X,Y,0,1) RGBA texels.  Such formats have
        not proven popular.  In particular, they interact awkwardly with
        the pixel path and conventional texture environment modes.

        The (X,Y,0,1) convention, particularly with signed components,
        is nice for normal maps because a normalized vector can be
        formed by a shader program by computing sqrt(abs(1-X*X-Y*Y))
        for the Z component.  However, this niceness is mostly conceptual
        however since the same effect can be accomplished with swizzling
        as shown in this GLSL code:

            vec2 texLA  = texture2D(samplerLA, gl_TexCoord[0]).xw;
            vec3 normal = vec3(texLA.x,
                               texLA.y,
                               sqrt(abs(1-texLA.x*texLA.x-texLA.y*texLA.y)));

        The most important reason to make these new compressed formats
        show up identically to conventional luminance and luminance-alpha
        texels is to allow applications to seamlessly substitute
        the new compressed formats for existing GL_LUMINANCE and
        GL_LUMINANCE_ALPHA textures.  Alternative component arrangements
        would make it more cumbersome for existing applications to switch
        over luminance and luminance-alpha textures to these compressed
        formats.

    3)  Should luminance and luminance-alpha compression formats with
        signed components be introduced when the core specification
        lacked uncompressed luminance and luminance-alpha texture formats?

        RESOLVED:  Yes, signed luminance and luminance-alpha compression
        formats should be added.

        Signed luminance-alpha formats are suited for compressed normal
        maps.  Compressed normal maps may well be the dominant use of
        this extension.

        Unsigned luminance-alpha formats require an extra "expand normal"
        operation to convert [0,1] to [-1,+1].  Direct support for signed
        luminance-alpha formats avoids this step in a shader program.

    4)  Should there be a mix of signed luminance and unsigned alpha or
        vice versa?

        RESOLVED:  No.

        NV_texture_shader provided an internal format
        (GL_SIGNED_RGB_UNSIGNED_ALPHA_NV) with mixed signed and unsigned
        components.  The format saw little usage.  There's no reason to
        think a GL_SIGNED_LUMINANCE_UNSIGNED_ALPHA format would be any
        more useful or popular.

    5)  How are signed integer values mapped to floating-point values?

        RESOLVED:  A signed 8-bit two's complement value X is computed to
        a floating-point value Xf with the formula:

                 { X / 127.0, X > -128
            Xf = {
                 { -1.0,      X == -128

        This conversion means -1, 0, and +1 are all exactly representable,
        however -128 and -127 both map to -1.0.  Mapping -128 to -1.0
        avoids the numerical awkwardness of have a representable value
        slightly more negative than -1.0.

        This conversion is intentionally NOT the "byte" conversion listed
        in Table 2.9 for component conversions.  That conversion says: 

            Xf = (2*X + 1) / 255.0

        The Table 2.9 conversion is incapable of exactly representing
        zero.

    6)  How will signed components resulting from
        GL_COMPRESSED_SIGNED_LUMINANCE_LATC1_EXT and
        GL_COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT texture fetches
        interact with fragment coloring?

        RESOLVED:  The specification language for this extension is silent
        about clamping behavior leaving this to the core specification
        and other extensions.  The clamping or lack of clamping is left
        to the core specification and other extensions.

        For assembly program extensions supporting texture fetches
        (ARB_fragment_program, EXT_fragment_program, EXT_vertex_program3,
        etc.) or the OpenGL Shading Language, these signed formats will
        appear as expected with unclamped signed components as a result
        of a texture fetch instruction.

        If ARB_color_buffer_float is supported, its clamping controls
        will apply.

        NV_texture_shader extension, if supported, adds support for
        fixed-point textures with signed components and relaxed the
        fixed-function texture environment clamping appropriately.  If the
        NV_texture_shader extension is supported, its specified behavior
        for the texture environment applies where intermediate values
        are clamped to [-1,1] unless stated otherwise as in the case
        of explicitly clamped to [0,1] for GL_COMBINE.  or clamping the
        linear interpolation weight to [0,1] for GL_DECAL and GL_BLEND.

        Otherwise, the conventional core texture environment clamps
        incoming, intermediate, and output color components to [0,1].

        This implies that the conventional texture environment
        functionality of unextended OpenGL 1.5 or OpenGL 2.0 without
        using GLSL (and with none of the extensions referred to above)
        is unable to make proper use of the signed texture formats added
        by this extension because the conventional texture environment
        requires texture source colors to be clamped to [0,1].  Texture
        filtering of these signed formats would be still signed, but
        negative values generated post-filtering would be clamped to
        zero by the core texture environment functionality.  The
        expectation is clearly that this extension would be co-implemented
        with one of the previously referred to extensions or used with
        GLSL for the new signed formats to be useful.

    7)  Should a specific normal map compression format be added?

        RESOLVED:  No.

        It's probably short-sighted to design a format just for normal
        maps.  Indeed, NV_texture_shader added a GL_SIGNED_HILO_NV
        format with exactly the kind of "hemisphere remap" useful for
        normal maps and the format went basically unused.  Instead,
        this extension provides the mechanism for compressed normal maps
        based on the more conventional luminance-alpha format.

        The GL_COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT and
        GL_COMPRESSED_SIGNED_LUMINANCE_ALPHA_LATC2_EXT formats are
        sufficient for normal maps with additional shader instructions
        used to generate the 3rd component.

    8)  Should uncompressed signed luminance and luminance-alpha formats
        be added by this extension?

        RESOLVED:  No, this extension is focused on just adding compressed
        texture formats.

        The NV_texture_shader extension adds such uncompressed signed
        texture formats.  A distinct multi-vendor extension for signed
        fixed-point texture formats could provide all or a subset of
        the signed fixed-point uncompressed texture formats introduced
        by NV_texture_shader.

    9)  What compression ratios does this extension provide?

        The LATC1 formats are 8 bytes (64 bits) per 4x4 pixel block.
        A 4x4 block of GL_LUMINANCE8 data requires 16 bytes (1 byte
        per texel).  This is a 2-to-1 compression ratio.

        The LATC2 formats are 16 bytes (128 bits) per 4x4 pixel block.
        A 4x4 block of GL_LUMINANCE8_ALPHA8 data requires 32 bytes
        (2 bytes per texel).  This is again a 2-to-1 compression ratio.

        In contrast, the comparable compression ratio for the S3TC
        formats is 4-to-1.

        Arguably, the lower compression ratio allows better compression
        quality particularly because the LATC formats compress each
        component separately.

    10) How do these new formats compare with the existing GL_LUMINANCE4,
        GL_LUMINANCE4_ALPHA4, and GL_LUMINANCE6_ALPHA2 internal formats?

        RESOLVED:  The existing GL_LUMINANCE4, GL_LUMINANCE4_ALPHA4,
        and GL_LUMINANCE6_ALPHA2 internal formats provide a similar
        2-to-1 compression ratio but mandate a uniform quantization
        for all components.  In contrast, this extension provides a
        compression format with 3-bit quantization over a specifiable
        min/max range that can vary per 4x4 texel tile.

        Additionally, many OpenGL implementations do not natively support
        the GL_LUMINANCE4, GL_LUMINANCE4_ALPHA4, and GL_LUMINANCE6_ALPHA2
        internal formats but rather silently promote these formats
        to store 8 bits per component, thereby eliminating any
        storage/bandwidth advantage for these formats.

    11) Does this extension require EXT_texture_compression_s3tc?

        RESOLVED:  No.

        As written, this specification does not rely on wording of the
        EXT_texture_compression_s3tc extension.  For example, certain
        discussion added to Sections 3.8.2 and 3.8.3 is quite similar
        to corresponding EXT_texture_compression_s3tc language.

    12) Should anything be said about the precision of texture filtering
        for these new formats?

        RESOLVED:  No precision requirements are part of the specification
        language since OpenGL extensions typically leave precision
        details to the implementation.

        Realistically, at least 8-bit filtering precision can be expected
        from implementations (and probably more).

    13) Should these formats be allowed to specify 3D texture images
        when NV_texture_compression_vtc is supported?

        RESOLVED: The NV_texture_compression_vtc stacks 4x4 blocks into
        4x4x4 bricks.  It may be more desirable to represent compressed
        3D textures as simply slices of 4x4 blocks.

        However the NV_texture_compression_vtc extension expects
        data passed to the glCompressedTexImage commands to be "bricked"
        rather than blocked slices.

    14) Why is GL_NV_texture_compression_latc also listed in the Name Strings
        section?

        The very first GeForce 8800 driver shipped with the extension
        designated as NV before EXT-ization with S3 was agreed.
        Subsequent NVIDIA drivers will rename the extension to its EXT
        name only.

    15) Should the the generic formats
        GL_COMPRESSED_LUMINANCE and GL_COMPRESSED_LUMINANCE_ALPHA
        correspond to COMPRESSED_LUMINANCE_LATC1_EXT and
        COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT respecitively when this
        extension is supported?

        RESOLVED:  Yes.  While no generic compression is strictly
        required for an implementation and there might exist superior
        compression schemes for luminance and luminance-alpha textures
        in the future, an application should reasonably expect that an
        implementation that supports EXT_texture_compression_latc will
        also use these formats for the generic compressed luminance and
        luminance-alpha formats.

        The COMPRESSED_LUMINANCE_LATC1_EXT and
        COMPRESSED_LUMINANCE_ALPHA_LATC2_EXT are generic enough in their
        respective luminance and luminance-alpha behavior that these
        compression formats are acceptable generic compressed formats
        for luminance and luminance-alpha generic compressed formats.

    16) Should the GL_NUM_COMPRESSED_TEXTURE_FORMATS and
        GL_COMPRESSED_TEXTURE_FORMATS queries return the LATC formats?

        RESOLVED:  No.

        The OpenGL 2.1 specification says "The only values returned
        by this query [GL_COMPRESSED_TEXTURE_FORMATS"] are those
        corresponding to formats suitable for general-purpose usage.
        The renderer will not enumerate formats with restrictions that
        need to be specifically understood prior to use."

        Historically, OpenGL implementation have advertised the RGB and
        RGBA versions of the S3TC extensions compressed format tokens
        through this mechanism.

        The specification is not sufficiently clear about what "suitable
        for general-purpose usage" means.  Historically that seems to mean
        unsigned RGB or unsigned RGBA.  The DXT1 format supporting alpha
        (GL_COMPRESSED_RGBA_S3TC_DXT1_EXT) is not exposed in the list (at
        least for NVIDIA drivers) because the alpha is always 1.0 expect
        when it is 0.0 when RGB is required to be black.  NVIDIA's even
        limits itself to true linear RGB or RGBA formats, specifically
        not including EXT_texture_sRGB's sRGB S3TC compressed formats.

        Adding luminance and luminance-alpha texture formats (and
        certainly signed versions of luminance and luminance-alpha
        formats!) invites potential comptaibility problems with old
        applications using this mechanism since old applications are
        unlikely to expect non-RGB or non-RGBA formats to be advertised
        through this mechanism.  However no specific misinteractions
        with old applications is known.

        Applications that seek to use the LATC formats should do so
        by looking for this extension's name in the string returned by
        glGetString(GL_EXTENSIONS) rather than
        what GL_NUM_COMPRESSED_TEXTURE_FORMATS and
        GL_COMPRESSED_TEXTURE_FORMATS return.

Implementation Note

    This extension allows TexSubImage2D and CompressedTexSubImage2D to perform
    partial updates to compressed images, but generally requires that the
    updated area be aligned to 4x4 block boundaries.  If the width or height
    is not a multiple of four, there will be 4x4 blocks at the edge of the
    image that contain "extra" texels that are not part of the image.  This
    spec has an exception allowing edits that partially cover such blocks as
    long as the edit covers all texels in the block belonging to the image.
    For example, in a 2D texture of size 70x50, it is legal to update the
    single partial block covering the four texels from (68,48) to (69,49) by
    setting (<xoffset>, <yoffset>) to (68,48) and <width> and <height> to 2.

    This specification derived some of its language from the
    EXT_texture_compression_s3tc specification.  When that extension was
    originally written, non-bordered textures were required to have widths and
    heights that were powers of two.  Therefore, the only cases where partial
    blocks could occur were if the width or height of the texture image was
    one or two.  The original spec language allowed partial block edits only
    if the width or height of the region edited was equal to the full texture
    size.  That language didn't handle cases such as the 70x50 example above.

    This specification was updated in April, 2009 to allow such edits.
    Multiple OpenGL implementers correctly implemented the original
    restriction, and partial edits that include partially covered tiles will
    result in INVALID_OPERATION errors on older drivers.


Revision History

    Revision 1.1, April 24, 2007: mjk
        -  Add caveat about how signed LA decompression happens when
           lum0 equals -127 and lum1 equals -128.  This caveat matches
           a decoding allowance in DirectX 10.

    Revision 1.2, January 21, 2008: mjk
        -  Add issues #15 and #16.

    Revision 1.3, April 14, 2009: pbrown 
        - Add interaction with non-power-of-two textures from OpenGL 2.0 /
          ARB_texture_non_power_of_two.  Allow CompressedTexSubImage2D to
          perform edits that include partial tiles at the edge of the image as
          long as the specified width/height parameters line up with the edge.
          Thanks to Emil Persson for finding this issue.

    Revision 2, July 30, 2017: dgkoch
	- replace references to COMPRESSED_LUMINANCE_LACT1_EXT with references
          to the correct spelling COMPRESSED_LUMINANCE_LATC1_EXT.