Name

    NV_shader_atomic_float64

Name Strings

    GL_NV_shader_atomic_float64

Contact

    Kedarnath Thangudu, NVIDIA Corporation (kthangudu 'at' nvidia.com)

Contributors

    Pat Brown, NVIDIA

Status

    Shipping in NVIDIA release 367.XX drivers and up.

Version

    Last Modified Date:         October 15, 2014
    NVIDIA Revision:            1

Number

    OpenGL Extension #488

Dependencies

    This extension is written against the OpenGL 4.5 (Compatibility Profile)
    Specification.

    This extension is written against version 4.50 (revision 3) of the OpenGL
    Shading Language Specification.

    This extension requires ARB_gpu_shader_fp64 or NV_gpu_program_fp64.

    This extension interacts with NV_shader_buffer_store, NV_gpu_shader5,
    ARB_shader_storage_buffer_object, and ARB_compute_shader.

    This extension interacts with NV_gpu_program5.

Overview

    This extension provides GLSL built-in functions and assembly opcodes
    allowing shaders to perform atomic read-modify-write operations to buffer
    or shared memory with double-precision floating-point components.  The set
    of atomic operations provided by this extension is limited to adds and
    exchanges. Providing atomic add support allows shaders to atomically
    accumulate the sum of double-precision floating-point values into buffer
    memory across multiple (possibly concurrent) shader invocations.

    This extension provides GLSL support for atomics targeting double-precision
    floating-point pointers (if NV_gpu_shader5 is supported).
    Additionally, assembly opcodes for these operations are also provided if
    NV_gpu_program5 is supported.

New Procedures and Functions

    None.

New Tokens

    None.

Additions to the OpenGL 4.5 (Compatibility Profile) Specification

    None.

Additions to the AGL/GLX/WGL Specifications

    None.

GLX Protocol

    None.

Modifications to the OpenGL Shading Language Specification, Version 4.50
(revision 3)

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

      #extension GL_NV_shader_atomic_float64 : <behavior>

    where <behavior> is as specified in section 3.3.

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

      #define GL_NV_shader_atomic_float64         1

    Modify Section 8.11, Atomic Memory Functions (p. 172)

    (add to "atomicAdd" table cell, p. 173)

      double atomicAdd(coherent inout double mem, double data)

    (add to "atomicExchange" table cell, p. 173)

      double atomicExchange(coherent inout double mem, double data)


Dependencies on NV_shader_buffer_store, NV_gpu_shader5,
ARB_shader_storage_buffer_object, and ARB_compute_shader

    If NV_shader_buffer_store and NV_gpu_shader5 are supported, the following
    functions should be added to the "Section 8.Y, Shader Memory Functions"
    language in the NV_shader_buffer_store specification:

      double atomicAdd(double *address, double data);
      double atomicExchange(double *address, double data);

    If ARB_shader_storage_buffer_object or ARB_compute_shader are supported,
    make similar edits to the functions documented in the
    ARB_shader_storage_buffer object extension.

    These functions are available if and only if GL_NV_shader_atomic_float64 is
    enabled via the "#extension" directive.

Dependencies on NV_gpu_program5

    If NV_gpu_program5 is supported and "OPTION NV_shader_atomic_float64" is
    specified in an assembly program, "F64" should be allowed as a storage
    modifier to the ATOM instruction for the atomic operations "ADD" and
    "EXCH".

    (Add to "Section 2.X.6, Program Options" of the NV_gpu_program4 extension,
    as extended by NV_gpu_program5:)

      + Double-precision Floating-Point Atomic Operations (NV_shader_atomic_float64)

      If a program specifies the "NV_shader_atomic_float64" option, it may use
      "F64" storage modifier with the "ATOM" opcode to perform atomic double-
      precision floating-point add or exchange operations.

    (Add to the table in "Section 2.X.8.Z, ATOM" in NV_gpu_program5:)

      atomic     storage
      modifier   modifiers                 operation
      --------   -----------------------   ---------------------------------
       ADD       U32, S32, U64, F32, F64   compute a sum
                 F16X2, F16X4
       ...
       EXCH      U32, S32, U64, F32, F64   exchange memory with operand
                 F16X2, F16X4

    Note:
      Storage modifier U64 is provided by NV_shader_atomic_int64
      Storage modifier F32 is provided by NV_shader_atomic_float
      Storage modifiers F16X2 and F16X4 are provided by NV_shader_atomic_fp16_vector

Errors

    None.

New State

    None.

New Implementation Dependent State

    None.

Issues

    (1) What double-precision floating-point targets are supported for
    atomic operations?

      RESOLVED: This extension only supports atomic operations on double-
      precision floating-point buffer memory. Atomic operation on double-
      precision texture memory are not supported since OpenGL provides
      no pixel/texture formats with double-precision components.

    (2) What atomic operations should we support for double-precision
    floating-point targets?

      RESOLVED:  Double-precision floating-point atomic addition is the main
      functionality targeted by this extension.  We provide exchanges because
      the operation needs no special hardware support.

      We chose not to provide support for bitwise operations (AND/OR/XOR);
      it's possible to support these by casting a pointer or aliasing an image
      if required.  Minimum, maximum, and compare-and-swap make sense, but the
      underlying atomic hardware targeted by this extension does not support
      floating-point comparisons.

Revision History

    Revision 1
      - Internal revisions.