NV_depth_nonlinear

GL_NV_depth_nonlinear
EGL_NV_depth_nonlinear

Gary King, NVIDIA Corporation (gking 'at' nvidia.com)

Copyright NVIDIA Corporation, 2005 - 2007.

NVIDIA Proprietary

Last Modified: 2007/03/20
NVIDIA Revision: 1.0

EGL Extension #18
OpenGL ES Extension #73

Written based on the wording of the OpenGL 2.0 Specification and
EGL 1.2 Specification.

Requires EGL 1.1.

Requires OpenGL-ES 1.0.

OES_framebuffer_object affects the wording of this specification.

Due to the perspective divide, conventional integer Z-buffers have
a hyperbolic distribution of encodings between the near plane
and the far plane.  This can result in inaccurate depth testing,
particularly when the number of depth buffer bits is small
and objects are rendered near the far plane.

Particularly when the number of depth buffer bits is limited
(desirable and/or required in low-memory environments), artifacts
due to this loss of precision may occur even with relatively
modest far plane-to-near plane ratios (e.g., greater than 100:1).

Many attempts have been made to provide alternate encodings for
Z-buffer (or alternate formulations for the stored depth) to
reduce the artifacts caused by perspective division, such as
W-buffers, Z-complement buffers and floating-point 1-Z buffers.

This extension adds a non-linear encoded Z buffer to OpenGL,
which can improve the practically useful range of, e.g. 16-bit
depth buffers by up to a factor of 16, greatly improving depth
test quality in applications where the ratio between the near
and far planes can not be as tightly controlled.

NVIDIA Proprietary

None

Accepted as a valid sized internal format by all functions accepting
sized internal formats with a base format of DEPTH_COMPONENT

    DEPTH_COMPONENT16_NONLINEAR_NV     0x8E2C

Accepted by the <attrib_list> parameter of eglChooseConfig and
eglCreatePbufferSurface, and by the <attribute> parameter of
eglGetConfigAttrib

    EGL_DEPTH_ENCODING_NV              0x30E2

Accepted as a value in the <attrib_list> parameter of eglChooseConfig
and eglCreatePbufferSurface, and returned in the <value> parameter
of eglGetConfigAttrib

    EGL_DEPTH_ENCODING_NONE_NV         0
    EGL_DEPTH_ENCODING_NONLINEAR_NV    0x30E3

Add the following line to table 3.16 (p. 154)

+--------------------------------+-----------------+------+
|      Sized Internal Format     |  Base Internal  |  D   |
|                                |  Format         | Bits |
+--------------------------------+-----------------+------+
| DEPTH_COMPONENT16_NONLINEAR_NV | DEPTH_COMPONENT |  16  |
+--------------------------------+-----------------+------+

Add the following line to table 3.1 (p. 14)

+--------------------------+------+---------------------------------------+
|         Attribute        | Type | Notes                                 |
+--------------------------+------+---------------------------------------+
|   EGL_DEPTH_ENCODING_NV  | enum | Type of depth-buffer encoding employed|
+--------------------------+------+---------------------------------------+

Modify the description of the depth buffer in Section 3.4 (p. 15)

"The depth buffer is used only by OpenGL ES.  It contains fragment depth 
(Z) information generated during rasterization.  EGL_DEPTH_SIZE indicates 
the depth of this buffer in bits, and EGL_DEPTH_ENCODING_NV indicates which
alternate depth-buffer encoding (if any) should be used.  Legal values for
EGL_DEPTH_ENCODING_NV are: EGL_DONT_CARE, EGL_DEPTH_ENCODING_NONE_NV and
EGL_DEPTH_ENCODING_NONLINEAR_NV."

Add the following line to table 3.4 (p. 20)

+-----------------------+---------------+-----------+-------+----------+
|       Attribute       |     Default   | Selection |  Sort |   Sort   |
|                       |               |  Criteria | Order | Priority |
+-----------------------+---------------+-----------+-------+----------+
| EGL_DEPTH_ENCODING_NV | EGL_DONT_CARE |   Exact   |  None |     -    |
+-----------------------+---------------+-------------------+----------+

None