Name

    ARB_ES3_1_compatibility

Name Strings

    GL_ARB_ES3_1_compatibility

Contact

    Piers Daniell, NVIDIA Corporation (pdaniell 'at' nvidia.com)

Contributors

    Daniel Koch, NVIDIA Corporation (dkoch 'at' nvidia.com)

    and contributors to the OpenGL ES 3.1 specification

Notice

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

Status

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

Version

    Last Modified Date:         September 19, 2014
    Revision:                   15

Number

    ARB Extension #159

Dependencies

    OpenGL 4.4, ARB_ES2_compatibility, ARB_ES3_compatibility are required.

    This extension is written against The OpenGL 4.4 (Compatibility Profile)
    specification.

Overview

    This extension adds support for features of OpenGL ES 3.1 that are
    missing from OpenGL 4.4. Enabling these features will ease the process
    of porting applications from OpenGL ES 3.1 to OpenGL.

    In particular this adds the following features:

    - a new MemoryBarrierByRegion API which is potentially more efficient
    for specific localized memory access patterns.

    - increases the minimum required size of SSBOs to 2^27 (128 MB).

    - support for GLSL ES version 310 (ie #version 310 es).

    - a new GLSL built-in function, imageAtomicExchange, which performs atomic
    exchanges on r32f floating point images.

    - a new GLSL built-in fragment shader input, gl_HelperInvocation, that
    identifies whether the current fragment shader input is a helper
    invocation.  Fragment shader code can use this variable to skip performing
    operations that are useless or potentially dangerous for helper
    invocations.

    - a new GLSL built-in constant for the maximum supported samples:
    gl_MaxSamples.

    - a number of new GLSL built-in constants mirroring the API limits for
    image uniforms: gl_Max*ImageUniforms, gl_MaxCombinedShaderOutputResources.

    - new GLSL built-in functions which extend mix() to select between int,
    uint, and bool components.

    - add the "coherent" qualifier to all memory variables taken by the GLSL
    built-in atomic* and imageAtomic* functions.

New Procedures and Functions

    void MemoryBarrierByRegion(bitfield barriers);

New Tokens

    None

Additions to Chapter 1 of the OpenGL 4.4 (Compatibility Profile)
Specification (Introduction)

    Modify section 1.3.2 (OpenGL ES) to include mention of OpenGL ES 3.1:

    OpenGL ES is a ... well-defined subsets of OpenGL.  OpenGL ES version
    1.1 implements a subset of the OpenGL 1.5 fixed-function API, OpenGL
    ES 2.0 implements a subset of the OpenGL 2.0 shader-based API, OpenGL
    ES 3.0 implements a subset of OpenGL 3.3, and OpenGL ES 3.1 implements
    a subset of OpenGL 4.3. OpenGL ES versions also include ...

    OpenGL and OpenGL ES are developed ...

    OpenGL implementations supporting this extension include functionality
    initially defined in OpenGL ES 3.1, for increased compatibility between
    OpenGL and OpenGL ES implementations.


    Modify section 1.3.3 (OpenGL ES Shading Language) to include mention
    of version 3.10:

    The Specification should also be read together with companion documents
    titled "The OpenGL ES Shading Language". Versions 1.00, 3.00 and 3.10
    should be read. These documents define versions of the OpenGL ES Shading
    Language designed for implementations of OpenGL ES 2.0, 3.0 and 3.1,
    respectively, but also supported by OpenGL implementations. References ...

    OpenGL implementations supporting this extensions are guaranteed to
    support versions 1.00, 3.00 and 3.10 of the OpenGL ES Shading Language.


Additions to Chapter 7 of the OpenGL 4.4 (Compatibility Profile)
Specification (Programs and Shaders)

    Modify Section 7.12.2, "Shader Memory Access Synchronization", (p. 148)

    (Append to the end of the section)

    "The command

        void MemoryBarrierByRegion(bitfield barriers)

    behaves as described above for MemoryBarrier, with two differences:

    First, it narrows the region under consideration so that only reads/writes
    of prior fragment shaders that are invoked for a smaller region of the
    framebuffer will be completed/reflected prior to subsequent reads/write of
    following fragment shaders.  The size of the region is implementation
    dependent and may be as small as one framebuffer pixel.
    
    Second, it only applies to memory transactions that may be read by or
    written by a fragment shader.  Therefore, only the barrier bits
      - ATOMIC_COUNTER_BARRIER_BIT
      - FRAMEBUFFER_BARRIER_BIT
      - SHADER_IMAGE_ACCESS_BARRIER_BIT
      - SHADER_STORAGE_BARRIER_BIT
      - TEXTURE_FETCH_BARRIER_BIT
      - UNIFORM_BARRIER_BIT
    are supported.

    When barriers is ALL_BARRIER_BITS, shader memory accesses will be
    synchronized relative to all these barrier bits, but not to other barrier
    bits specific to MemoryBarrier. This implies that reads/writes for
    scatter/gather-like algorithms may or may not be completed/reflected after
    a MemoryBarrierByRegion command. However, for uses such as deferred
    shading, where a linked list of visible surfaces with the head at a
    framebuffer address may be constructed, and the entirety of the list is
    only dependent on previous executions at that framebuffer address,
    MemoryBarrierByRegion may be significantly more efficient than
    MemoryBarrier."


Additions to Chapter 9 of the OpenGL 4.4 (Compatibility Profile)
Specification (Framebuffers and Framebuffer Objects)

    Modify Section 9.2.3 (Framebuffer Object Queries) in the second
    paragraph describing GetFramebufferAttachmentParameteriv to add BACK as
    a valid <attachment>:

    "If the default framebuffer is bound to <target>, then <attachment> must
    be one of FRONT_LEFT, FRONT_RIGHT, BACK, BACK_LEFT, or BACK_RIGHT,
    identifying a color buffer; ... identifying the stencil buffer. Since
    this command can only query a single framebuffer attachment, BACK is
    equivalent to BACK_LEFT."

    Modify the first paragraph of Section 9.4.5 (Effects of
    Framebuffer State on Framebuffer Dependent Values)
    
    "The values of the state variables listed in table 23.85 may change when
    a change is made to the current framebuffer binding, to the state of the
    currently bound framebuffer object, or to an image attached to that
    framebuffer object. Most such state is dependent on the draw framebuffer
    (DRAW_FRAMEBUFFER_BINDING), but IMPLEMENTATION_COLOR_READ_TYPE and
    IMPLEMENTATION_COLOR_READ_FORMAT are dependent on the read framebuffer
    (READ_FRAMEBUFFER_BINDING)."


Additions to Chapter 15 of the OpenGL 4.4 (Compatibility Profile)
Specification (Programmable Fragment Processing)

    Modify Section 15.2.2 (Shader Inputs), (p. 530, 5th paragraph)

    "The built-in variable gl_SampleMaskIn is an integer array...
    The number of elements in the array is ceil(gl_MaxSamples/32), where
    gl_MaxSamples is the value of MAX_SAMPLES, the maximum number of
    color samples supported by the implementation.  Bit <n> ... "

    Modify Section 15.2.3 (Shader Outputs), (p. 532, 2nd paragraph)

    "The built-in integer array gl_SampleMask can be used ...
    The number of elements in the arrays is ceil(gl_MaxSamples/32), where
    gl_MaxSamples is the value of MAX_SAMPLES, the maximum number of
    color samples supported by the implementation. If bit <n> ... "


Additions to Chapter 17 of the OpenGL 4.4 (Compatibility Profile)
Specification (Writing Fragments and Samples to the Framebuffer)

    Modify Section 17.4.1 (Selecting Buffers for Writing), (p. 572,
    2nd paragraph)
    
    "If the GL is bound to the default framebuffer, then each of the
    constants must be one of the values listed in table 17.6 or the
    special value BACK. When the value BACK is used then <n> must
    be 1 and color values are written into the left buffer for
    single-buffered contexts, or into the back left buffer for double-
    buffered contexts. An INVALID_OPERATION error is generated if
    <bufs> contains the value BACK and <n> does not equal 1."


Additions to Chapter 18 of the OpenGL 4.4 (Compatibility Profile)
Specification (Drawing, Reading, and Copying Pixels)

    Modify Section 18.2.1 (Selecting Buffers for Reading), (p. 589)
    
    Remove the last paragraph that describes the
    IMPLEMENTATION_COLOR_READ_FORMAT and IMPLEMENTATION_COLOR_READ_TYPE
    queries and errors.
    
    Modify Section 18.2.2 (ReadPixels), (p. 589)
    
    Add the following paragraph at the end of the section:
    
    "The preferred parameters for <format> and <type> may be determined by
    calling GetIntegerv with the symbolic constants
    IMPLEMENTATION_COLOR_READ_FORMAT and IMPLEMENTATION_COLOR_READ_TYPE,
    respectively. The implementation chosen format may vary depending on the
    format of the selected read buffer of the currently bound read
    framebuffer.
    
    An INVALID_OPERATION error is generated by GetIntegerv if pname is
    IMPLEMENTATION_COLOR_READ_FORMAT or IMPLEMENTATION_COLOR_READ_TYPE and
    any of:
     - the read framebuffer is not framebuffer complete
     - the read framebuffer is a framebuffer object, and the selected read buffer
       (see section 18.2.1) has no image attached
     - the selected read buffer is NONE
    "

Additions to Chapter 22 of the OpenGL 4.4 (Compatibility Profile)
Specification (Context State Queries)

    Modify Section 22.2 (Pointer and String Queries), (p. 648)

    "If <name> is SHADING_LANGUAGE_VERSION, ...
    Version strings "100", "300 es" and "310 es" correspond to the OpenGL
    ES Shading Language versions 100, 300 and 310, respectively."


Additions to the OpenGL Shading Language

    Including the following line in a shader can be used to control the
    language features described in this extension:

        #extension GL_ARB_ES3_1_compatibility : <behavior>

    where <behavior> is as specified in section 3.3.

    New preprocessor #defines are added to the OpenGL Shading Language:

        #define GL_ARB_ES3_1_compatibility 1


Additions to Chapter 3 of the OpenGL Shading Language 4.40.8 Specification
(Basics)

    Modify the paragraph at the bottom of page 16 in Section 3.3
    (Preprocessor) as follows:

    "... Shaders that specify #version 100 will be treated as targeting
    version 1.00 of the OpenGL ES Shading Language. Shaders that specify
    #version 300 will be treated as targeting version 3.00 of the OpenGL
    ES Shading Language. Shaders that specify #version 310 will be treated
    as targeting version 3.10 of the OpenGL ES Shading Language. ..."

    Modify the 3rd paragraph at the top of p16 as follows:

    A <profile> argument ...
    If version 300 or 310 is specified, the profile argument is not optional
    and must be "es", or a compile-time error results. The Language
    Specification ...


Additions to Chapter 4 of the OpenGL Shading Language 4.40.8 Specification
(Variables and Types)

    Modify Section 4.7.2 (Precision Qualifiers)
    
    (Replace the first sentence to reference all opaque types instead of just
    samplers)
     
    "Any floating point, integer, or opaque-type declaration can have the
    type preceded by one of these precision qualifiers: ..."
    
    Modify Section 4.7.3 (Default Precision Qualifiers)
    
    (Replace the second sentence to reference all opaque types and not just
    samples)
    
    "The type field can be either int or float, any of the opaque types, and
    the precision-qualifier can be lowp, mediump, or highp. ..."


Additions to Chapter 7 of the OpenGL Shading Language 4.40.8 Specification
(Built-in Variables)

    Modify Section 7.1, Built-In Language Variables (p. 122)

    (Add to list of fragment language built-in variables on page 122)

        in bool gl_HelperInvocation;

    (Add after the description of gl_SamplePosition on page 127)

    "The value gl_HelperInvocation is true if the fragment shader invocation is
    considered a "helper invocation" and is false otherwise.  A "helper
    invocation" is a fragment-shader invocation that is created solely for the
    purposes of evaluating derivatives for use in non-helper fragment-shader
    invocations.  Such derivatives are computed implicitly in the built-in
    function texture() (see section 8.9 "Texture Functions"), and explicitly in
    the derivative functions in section 8.13.1 "Derivative Functions", for 
    example dFdx() and dFdy().

    Fragment shader helper invocations execute the same shader code as
    non-helper invocations, but will not have side effects that modify the
    framebuffer or other shader-accessible memory.  In particular:

      * Fragments corresponding to helper invocations are discarded when
        shader execution is complete, without updating the framebuffer.

      * Stores to image and buffer variables performed by helper invocations
        have no effect on the underlying image or buffer memory.

      * Atomic operations to image, buffer, or atomic counter variables
        performed by helper invocations have no effect on the underlying image
        or buffer memory.  The values returned by such atomic operations are
        undefined.

    Helper invocations may be generated for pixels not covered by a primitive
    being rendered.  While fragment shader inputs qualified with "centroid"
    are normally required to be sampled in the intersection of the pixel and
    the primitive, the requirement is ignored for such pixels since there is
    no intersection between the pixel and primitive.

    Helper invocations may also be generated for fragments that are covered by
    a primitive being rendered when the fragment is killed by early fragment
    tests (using the "early_fragment_tests" qualifier) or where the
    implementation is able to determine that executing the fragment shader
    would have no effect other than assisting in computing derivatives for
    other fragment shader invocations.

    The set of helper invocations generated when processing any set of
    primitives is implementation-dependent."

    Add to section 7.3 Built-in Constants

        const int gl_MaxSamples = 4;

        const int gl_MaxVertexImageUniforms = 0;
        const int gl_MaxFragmentImageUniforms = 8;
        const int gl_MaxComputeImageUniforms = 8;
        const int gl_MaxCombinedImageUniforms = 48;
        const int gl_MaxCombinedShaderOutputResources = 16;

Additions to Chapter 8 of the OpenGL Shading Language 4.40.8 Specification
(Built-in Functions)

    Modify Section 8.3, Common Functions

    (Additions to the table listing common built-in functions)

      Syntax                       Description
      ---------------------------  --------------------------------------------------
      genIType mix(genIType x,     Selects which vector each returned component comes
                   genIType y,     from. For a component of a that is false, the
                   genBType a)     corresponding component of x is returned. For a
      genUType mix(genUType x,     component of a that is true, the corresponding
                   genUType y,     component of y is returned.
                   genBType a)
      genBType mix(genBType x,
                   genBType y,
                   genBType a)

    Modify Section 8.11, Atomic Memory Functions

    (Modify the table on page 173 by adding "coherent" qualifier to all memory
     variables.)

      Syntax                                    Description
      ----------------------------------------  ---------------------------------------------------
      uint atomicAdd(coherent inout uint mem,   Computes a new value by adding the value of data to
                     uint data)                 the contents mem.
      int atomicAdd(coherent inout int mem,
                    int data)

      uint atomicMin(coherent inout uint mem,   Computes a new value by taking the minimum of the
                     uint data)                 value of data and the contents of mem.
      int atomicMin(coherent inout int mem,
                    int data)

      uint atomicMax(coherent inout uint mem,   Computes a new value by taking the maximum of the
                    uint data)                  value of data and the contents of mem.
      int atomicMax(coherent inout int mem,
                    int data)

      uint atomicAnd(coherent inout uint mem,   Computes a new value by performing a bit-wise
                     uint data)                 AND of the value of data and the contents of mem.
      int atomicAnd(coherent inout int mem,
                    int data)

      uint atomicOr(coherent inout uint mem,    Computes a new value by performing a bit-wise OR
                    uint data)                  of the value of data and the contents of mem.
      int atomicOr(coherent inout int mem,
                   int data)

      uint atomicXor(coherent inout uint mem,   Computes a new value by performing a bit-wise
                     uint data)                 EXCLUSIVE OR of the value of data and the
      int atomicXor(coherent inout int mem,     contents of mem.
                    int data)

      uint atomicExchange(                      Computes a new value by simply copying the value
          coherent inout uint mem,              of data.
          uint data)
      int atomicExchange(
          coherent inout int mem,
          int data)

      uint atomicCompSwap(                      Compares the value of compare and the contents of
          coherent inout uint mem,              mem. If the values are equal, the new value is given
          uint compare,                         by data; otherwise, it is taken from the original
          uint data)                            contents of mem.
      int atomicCompSwap(
          coherent inout int mem,
          int compare,
          int data)

    Modify Section 8.12, Image Functions

    (Modify the table section listing imageAtomic* on page 175 by adding
     the "coherent" qualifier to the image parameter and adding a new
     imageAtomicExchange function for float data.)

      Syntax                          Description
      ------------------------------  ---------------------------------------------------
      uint imageAtomicAdd(            Computes a new value by adding the value of data
          coherent IMAGE_PARAMS,      to the contents of the selected texel.
          uint data)
      int imageAtomicAdd(
          coherent IMAGE_PARAMS,
          int data)

      uint imageAtomicMin(            Computes a new value by taking the minimum of the
          coherent IMAGE_PARAMS,      value of data and the contents of the selected texel.
          uint data)
      int imageAtomicMin(
          coherent IMAGE_PARAMS,
          int data)

      uint imageAtomicMax(            Computes a new value by taking the maximum of the
          coherent IMAGE_PARAMS,      value data and the contents of the selected texel.
          uint data)
      int imageAtomicMax(
          coherent IMAGE_PARAMS,
          int data)

      uint imageAtomicAnd(            Computes a new value by performing a bit-wise
          coherent IMAGE_PARAMS,      AND of the value of data and the contents of the
          uint data)                  selected texel.
      int imageAtomicAnd(
          coherent IMAGE_PARAMS,
          int data)

      uint imageAtomicOr(             Computes a new value by performing a bit-wise OR
          coherent IMAGE_PARAMS,      of the value of data and the contents of the selected
          uint data)                  texel.
      int imageAtomicOr(
          coherent IMAGE_PARAMS,
          int data)

      uint imageAtomicXor(            Computes a new value by performing a bit-wise
          coherent IMAGE_PARAMS,      EXCLUSIVE OR of the value of data and the
          uint data)                  contents of the selected texel.
      int imageAtomicXor(
          coherent IMAGE_PARAMS,
          int data)

      uint imageAtomicExchange(       Computes a new value by simply copying the value of
          coherent IMAGE_PARAMS,      <data>. These functions support 32-bit signed and
          uint data);                 unsigned integer operands, and 32-bit float operands.
      int imageAtomicExchange(
          coherent IMAGE_PARAMS,
          int data);
      float imageAtomicExchange(
          coherent IMAGE_PARAMS,
          float data);

      uint imageAtomicCompSwap(       Compares the value of compare and the contents of
          coherent IMAGE_PARAMS,      the selected texel. If the values are equal, the new
          uint compare,               value is given by data; otherwise, it is taken from the
          uint data)                  original value loaded from the texel.
      int imageAtomicCompSwap(
          coherent IMAGE_PARAMS,
          int compare,
          int data)

Additions to the AGL/GLX/WGL Specifications

    None

Errors

    An INVALID_VALUE is generated by MemoryBarrier or MemoryBarrierByRegion
    if barriers is not the special value ALL_BARRIER_BITS and any of the
    unsupported bits are set.

New State

    (Modify Table 23.64 - Implementation Dependent Values)
    
    Remove IMPLEMENTATION_COLOR_READ_FORMAT and IMPLEMENTATION_COLOR_READ_TYPE.

    (Modify Table 23.76 - Implementation Dependent Aggregate Shader Limits)

    Increase the minimum required value of MAX_SHADER_STORAGE_BLOCK_SIZE to 2^27.

    (Modify Table 23.85 - Framebuffer Dependent Values)
    
    Add the following entries:
    
    Get Value                         Type  Get Command  Minimum Value  Description                     Sec
    --------------------------------  ----  -----------  -------------  ------------------------------  ----
    IMPLEMENTATION_COLOR_READ_TYPE    E     GetIntegerv  -              Implementation preferred pixel  18.2
                                                                        type*
    IMPLEMENTATION_COLOR_READ_FORMAT  E     GetIntegerv  -              Implementation preferred pixel  18.2
                                                                        format*

    "* Unlike most framebuffer-dependent state which is queried from the currently bound
      draw framebuffer, this state is queried from the currently bound read framebuffer."


Issues

    1) With this extension, is OpenGL 4.4 a complete superset of OpenGL ES 3.1?

    RESOLVED: No, there are things in OpenGL ES 3.1 that were once in OpenGL
    but have since been deprecated and removed. There is no plan to bring
    these back into core OpenGL. Here is a list of those things:

     - Application-generated object names.

     - Client vertex and index arrays.

     - Default vertex array object.

     - ALPHA, LUMINANCE and LUMINANCE_ALPHA pixel formats.

     - RED_BITS, GREEN_BITS, BLUE_BITS, DEPTH_BITS and STENCIL_BITS.

     - GENERATE_MIPMAP_HINT and automatic mipmap generation.

     - ALIASED_POINT_SIZE_RANGE query.

     - Line widths greater than 1.

     - Query of EXTENSIONS with GetString.

    In addition the error FRAMEBUFFER_INCOMPLETE_DIMENSIONS was in ES2 and the
    token remains in ES3 and above, but ES3 and above cannot generate this
    error. This has not been added to OpenGL core.

Revision History

    Rev.    Date    Author    Changes
    ----  --------  --------  -----------------------------------------------
    15    09/19/14  Jon Leech Remove FRONT as valid <attachment>.    

    14    09/18/14  Jon Leech Add FRONT and BACK as valid <attachment>s for 
                              GetFramebufferAttachmentParameteriv (Bug
                              12695).

    13    05/08/14  pdaniell  Allow precision qualifiers to be used with any
                              opaque type, and not just samplers.

    12    05/01/14  pdaniell  Add #extension support.

    11    04/17/14  pdaniell  Improve support for the
                              IMPLEMENTATION_COLOR_READ_FORMAT and
                              IMPLEMENTATION_COLOR_READ_TYPE queries to match
                              the OpenGL ES 3.1 spec.
    
    10    04/16/14  pdaniell  Add ES compatibility for calling DrawBuffers
                              with BACK.
    
    9     03/20/14  pdaniell  Update glMemoryBarrierByRegion language to
                              match latest OpenGL ES 3.1 spec.

    8     03/15/14  dkoch     Update overview to specifically itemize the
                              additions.

    7     03/13/14  pdaniell  Remove fixes for ARB_vertex_attrib_binding
                              which have now gone into an OpenGL 4.4 update.

    6     03/07/14  pdaniell  Update language for glMemoryBarrierByRegion to
                              the latest ES 3.1 spec.

    5     03/04/14  pdaniell  Add some more built-in constants for shader
                              image load store.

    4     03/04/14  dkoch     Remove unnecessary edits (Bug 11802).

    3     03/02/14  dkoch     More complete first draft.

    2     02/27/14  pdaniell  Complete the first draft.

    1     12/20/13  pdaniell  Initial draft with overview only.