blob: 04506c8259d42eb4b0a5294d14e7821b346ded7b [file] [log] [blame]
Name Strings
$Date: 1999/06/23 23:25:41 $ $Revision: 1.9 $
OpenGL 1.1 is required
SGIX_fbconfig is required
SGIX_fbconfig_float is recommended
EXT_histogram affects the definition of this extension
SGI_color_table affects the definition of this extension
SGIS_multisample affects the definition of this extension
This extension enhances the fbconfig types to support fbconfigs
(visuals) with extended range color data. The range of color data is
extended from [0,1] to [min,max], e.g., [-1.5,4.0]. Most OpenGL data
paths which clamp colors to [0,1] are modified to clamp colors to
[min,max]. The range is a fixed property of the fbconfig. Extended
range color is useful for imaging calculations and multipass shading
operations which often require signed data or increased dynamic range.
IP Status
Silicon Graphics has filed for patent protection for some of the
techniques described in this extension document.
* how does an application request the color range it needs to represent?
should the min/max values be interpreted specially in ChooseConfig
or should they be of a more limited usefulness exact match style?
They are interpreted uniquely. An fbconfig is chosen with the
tightest possible bound which meets or exceeds the range
indicated by the application.
* the spec does not distinguish between old-style [0,1] color
and extended range color (i.e. no attribute bits, just
the min and max values). Is this okay? NO!!!
Added EXTENDED_RANGE_SGI attribute to distinguish extended
range visuals from other visuals.
* float input pixel values and float colors automatically support the
extended range. Integer colors are still scaled to [0,1].
This means that GetIntegerv of an extended range color can not
represent an extended color range. Is this okay? YES.
* when the color range is extended this implies that accumulation buffers
and texture colors also support the extended range. Is this okay?
it has been suggested that accumulation stay as [-1,1] which would
imply that out of range results would leave the acbuf undefined?
* if a component isn't stored in the framebuffer (bits == 0) what should
range queries return. 0 seems obvious for R, G, and B, but what about
alpha. If there is no alpha value, should alpha be the max value?
no, alpha is 1.0
* treatment of alpha in general - whenever it would have been set to
1.0 it continues to be 1.0 not <max>
* do the ONE_MINUS* blend functions change? no. they are still 1.0-x
If the resulting value is negative it is clamped to <min> not 0.
Same is true for texture environment calculations.
* do transparent pixel values need to be embellished?
* floating point representations need some more clarification:
1. internal formats (need new extensions for these)
2. logic op (operates on mantissa or on the raw value?)
3. color table lookups (raw bits or clamped[0,1]
4. color indices - not supported
5. GLX_{RGBA}_SIZE - returns number of bits in the representation
* float versions of glXGetFBConfigAttribSGIX? YES!
* is clamp to [0,1] behavior necessary? it could be useful to have this
behavior where colors are used for blend factors in the texture
environment and color buffer blending operations.
* Is this extension SGI, SGIX, or SGIS? SGIS, because it does not
apply to all SGI platforms, however unlike SGIX, is intended for
wide acceptance.
New Procedures and Functions
New Tokens
Accepted by the <pname> parameters of GetBooleanv, GetIntegerv,
GetFloatv, and GetDoublev, the <attrib_list> parameter of
glXChooseFBConfigSGIX, and the <int_attrib_list> and
<flt_attrib_list> parameters of glXChooseFBConfigWithFltSGIX.
Accepted by the <attribute> parameter of glXGetFBConfigAttribSGIX
and glXGetFBConfigFltAttribSGIX, the <attrib_list> parameter of
glXChooseFBConfigSGIX, and the <int_attrib_list> and
<flt_attrib_list> parameters of glXChooseFBConfigWithFltSGIX:
Additions to Chapter 2 of the 1.1 Specification (OpenGL Operation)
Section 2.7 Vertex Specification
Versions of the Color command that take floating-point values
accept values nominally between <min> and <max>. <min> corresponds
to the minimum value while <max> corresponds to the maximum (machine
dependent) value that a component may take on in the framebuffer.
The three component variants of the color command set A to 1.0
Section 2.13 Colors and Coloring
Unsigned integer colors are still scaled to the range [0,1] and thus cannot
directly represent colors outside the range [0,1]. Signed integer and
floating-point colors can be used to represent colors outside the range
Section 2.13.6 (Clamping or Masking)
After lighting, RGBA colors are clamped to the range [min, max].
Section 2.13.9 (Final Color Processing)
For an RGBA color, each color component (which lies in [min,max]) is
converted (by rounding to nearest) to a value with m bits.
If the value is fixed-point we assume that the
representation used represents each value min+(max-min)k/(2^m - 1),
where k is in {0, 1, ..., 2^m - 1}, as
k (e.g. <max> is represented in binary as a string of all ones). m
must be at least as large as the number of bits in the corresponding
component of the framebuffer.
[XXX what should we assume if the value is not a fixed point value?]
[XXXare the rules for the msbs of unsigned color values matching the
framebuffer components true for color ranges outside [0,1]?]
Additions to Chapter 3 of the 1.1 Specification (Rasterization)
Section 3.6.3 (Rasterization of Pixel Rectangles)
Final Expansion to RGBA
if a group does not contain an A element, then A is added and set to 1.0.
If any of R, G, or B is missing from the group, each missing element is
assigned a value of 0.0.
RGBA to RGBA Lookup
First, each component is clamped to the range [0,1]. If there is a
table associated with each of the R, G, B, and A component elements:
and PIXEL_MAP_A_TO_A for A. Each element is multiplied by t
where t is an integer one less than the size of the corresponding table,
and, for each element, an address is found by rounding this value to
the nearest integer. For each element, the addressed value in the
corresponding table replaces the element.
If HISTOGRAM_EXT is enabled and the width of the table is non-zero, and
the pixel groups contain RGBA values, then indices Ri, Gi, Bi, and Ai
are derived from the red, green, blue, and alpha components of each
pixel group (without modifying these components) by clamping the
components to [0,1], multiplying each by t/(max-min) where t is one
less than the width of the histogram table, and rounding each to the
nearest integer. The color components used to index the histogram
are clamped, the original values are passed through unmodified to
the min/max stage.
Color Table
The color components of each group that are being replaced by table
values are converted to indices by clamping the components to
[0,1], multiplying each by t, where t is one less than
the width of the color table, and rounding each to the nearest
integer. The component value (R, G, B, or A) is then replaced by the
value in color table indicated in table E14.2, at the computed index.
Final Conversion
For RGBA components each element is clamped to [min,max]. The resulting
values are converted to fixed-point according to the rules given in
section 2.12.9 (Final Color Processing).
Section 3.8 (Texturing)
Each R, G, B, and A value so extracted is clamped to [min,max].
Each of the four values set by TEXTURE_BORDER_COLOR is clamped to lie
in [min,max].
Section 3.8.5 (Texture Environments and Texture Functions)
TEXTURE_ENV_COLOR is set to an RGBA color by providing four
single-precision floating-point values in the range [min,max] (values
outside this range are clamped to it).
R, G, B, and A values after being obtained from a supplied texture image,
are in the range [min, max].
[modifications to table 3.9???]
Section 3.9 (Fog)
Each component of Cf is clamped to [min, max] when specified.
Additions to Chapter 4 of the 1.1 Specification (Per-Fragment Operations
and the Frame buffer)
Color buffers consist of either unsigned integer color indices or
R, G, B, and , optionally A unsigned or signed integer values.
[XXXfloating point] Multisample Fragment Operations
If SAMPLE_ALPHA_TO_MASK_SGIS is enabled, the fragment alpha value used
for mask conversion is clamped to [0,1] before the mask is generated, but
the fragment alpha value is left unchanged. If SAMPLE_ALPHA_TO_ONE_SGIS
is enabled the fragment alpha is set to 1.0.
Section 4.1.3 (Alpha Test)
ref is clamped to lie in [min, max], and then converted to a fixed-point
value according to the rules given for an A component in section 2.13.9.
Section 4.2.3 (Clearing the Buffers)
void ClearColor(clampf r, clampf g, clampf b, clampf a);
sets the clear value for the color buffers in RGBA mode. Each of the
specified components is clamped to [min, max] and converted to fixed-point
according to the rules of section 2.12.9.
void ClearAccum(clampf r, clampf g, clampf b, clampf a);
takes four floating-point arguments that are the values, in order, to which
to set the R, G, B, and A values of the accumulation buffer (see the next
section). These values are clamped to the range [min2,max] where min2
= minimum(-1,<min>) when they are specified.
Section 4.2.4 (The Accumulation Buffer)
Accumulation buffer values are taken to be signed values in the range
[min2,max] where min2 = minimum(-1,<min>). Using ACCUM obtains the R, G,
B, and A components from the buffer current selected for
reading (section 4.3.2). Each component, considered as a fixed-point
values in [min,max] (see section 2.12.9) is converted to floating-point.
Each result is then multiplied by value.
The color components operated on by Accum must be clamped only if the
operation is RETURN. In this case, a value sent to the enabled color
buffers is first clamped to [min,max]. Otherwise, results are undefined
if the result of an operation on a color component is too large (in
magnitude) to be represented by the number of available bits.
4.3.2 (Reading Pixels)
Conversion to RGBA values
The R, G, and B (and possibly A) values form a group of elements. Each
element is taken to be a fixed-point value in [min,max] with m bits, where
m is the number of bits in the corresponding color component of the
selected buffer (see section 2.12.9)
Final Conversion
For a component, if the type is floating point then each component is first
clamped to [min, max]; if the type is signed integer, then the component is
clamped to the more restrictive of [-1,1] and [min,max], otherwise each
component is first clamped to [0,1]. Then the appropriate conversion
formula from Table 4.7 is applied to the component.
Additions to Chapter 5 of the 1.1 Specification (Special Functions)
Additions to Chapter 6 of the 1.1 Specification (State and State Requests)
Additions to the GLX Specification
[Added to the description of glXChooseFBConfigSGIX ]
If GLX_EXTENDED_RANGE_SGIS is specified as True, and if any of
GLX_MIN_ALPHA_SGIS, or GLX_MAX_ALPHA_SGIS are specified in <attrib_list>
then the value that follows indicates the minimum or maximum value
which can be represented in the OpenGL pipeline as a color. Any
OpenGL state which stores color values (e.g. current color, texture
data, the color buffer, accumulation buffer, aux buffers, etc) can
represent values between the minimum and maximum range. Integer
values for MIN and MAX attributes are cast to double precision
floating point, which is the internal representation of these
attributes. Non-integral values for these attributes may be
specified using glXChooseFBConfigWithFltSGIX, which is
defined in the SGIX_fbconfig_float GLX extension.
The visuals returned by glXChooseFBConfigSGIX are sorted on the
basis of color range if and only if GLX_EXTENDED_RANGE_SGIS is
specified as True. In that case, color range has a higher
precedence than GLX_DEPTH_BITS and a lower precedence than
Extended color range visuals can be specified for both displayable
(window) and non-displayable (pixmap) drawables. Extended range
visuals are displayed in a device-dependant manner, however for
traditional color displays they will typically be clamped to the
range [0,1].
[Added to Table 3]
Attribute Type Description
--------- ---- -----------
GLX_EXTENDED_RANGE_SGIS boolean true if visual is extended range
GLX_MIN_RED_SGIS double minimum value of red component
GLX_MAX_RED_SGIS double maximum value of red component
GLX_MIN_GREEN_SGIS double minimum value of green component
GLX_MAX_GREEN_SGIS double maximum value of green component
GLX_MIN_BLUE_SGIS double minimum value of blue component
GLX_MAX_BLUE_SGIS double maximum value of blue component
GLX_MIN_ALPHA_SGIS double minimum value of alpha component
GLX_MAX_ALPHA_SGIS double maximum value of alpha component
[Added to Table 4]
Attribute Default Sorting Criterion
--------- ------- -----------------
GLX_MIN_RED_SGIS 0.0 range
GLX_MAX_RED_SGIS 1.0 range
range - configs with a [MIN,MAX] range that meets or exceeds
the specified range are returned, with precedence given to
tighter bounds (that is, less negative MIN and less positive
GLX Protocol
Dependencies on EXT_histogram
If EXT_histogram is not implemented, then the references to
GetHistogramEXT and GetMinmaxEXT in this file are invalid, and should be
Dependencies on SGI_color_table
If SGI_color_table is not implemented, then the references to
ColorTableSGI and GetColorTableSGI in this file are invalid, and should
be ignored.
Dependencies on SGIS_multisample
If SGIS_multisample is not implemented, then the references to
and should be ignored.
New State
New Implementation Dependent State
Get Value Get Command Type Minimum Value Maximum Value
--------- ----------- ---- ------------- -------------
MIN_RED_SGIS GetFloatv R - 0.0
MAX_RED_SGIS GetFloatv R+ 1.0 -
MIN_GREEN_SGIS GetFloatv R - 0.0
MAX_GREEN_SGIS GetFloatv R+ 1.0 -
MIN_BLUE_SGIS GetFloatv R - 0.0
MAX_BLUE_SGIS GetFloatv R+ 1.0 -
MIN_ALPHA_SGIS GetFloatv R - 0.0
MAX_ALPHA_SGIS GetFloatv R+ 1.0 -