conformance.html | searchcode

/doc/conformance.html

https://github.com/excid3/chromium
HTML | 243 lines | 215 code | 24 blank | 4 comment | 0 complexity | f8491e6dba692e278cf7eaf367ed7d47 MD5 | raw file

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<!-- Copyright (c) 2001, Stanford University
     All rights reserved.

	 See the file LICENSE.txt for information on redistributing this software.  -->
<html>

<head>
<meta http-equiv="Content-Type" content="text/html; charset=windows-1252">
<meta name="GENERATOR" content="Microsoft FrontPage 4.0">
<meta name="ProgId" content="FrontPage.Editor.Document">
<title>New Page 1</title>
<base target="_self">
<link rel="stylesheet" type="text/css" href="chromium.css">
</head>

<body bgcolor="#FFFFFF">

<h1 class="PageHeading">Chromium OpenGL Conformance</h1>
<p>
Chromium strives to be a drop-in replacement for a conventional OpenGL
implementation.  In order to do so, Chromium must conform to the
OpenGL specification.  In some areas, this is very difficult or
impossible to achieve.  This section describes the ways in which
Chromium differs from OpenGL.
</p>

<h2 class="PageSubHeading">Conformance Issues</h2>

<p>
The following OpenGL features are not implemented in Chromium at this
time:
</p>
<ul>
<li>OpenGL 1.2 imaging functions related to histogram, min/max,
convolution and colortables.</li>
<li>To support Feedback & Selection, you need to use the feedbackSPU.
If you don't, these features won't work.</li>
</ul>

<p>
Display lists aren't fully conformant.  There are several issues:
</p>

<ul>
<li>When using the tilesort SPU, glCopyTexImage (and similar) commands
can't be implemented correctly since the framebuffer is tiled and the
individual renderers can't access their peers.
</li>
<li>GL_COMPILE_AND_EXECUTE mode may not work reliably, specifically
if there is a glGet* call between glNewList and glEndList which gets
state that was set by a previous command compiled/executed inside the
display list.
</li>
</ul>

<p>
glReadPixels and glCopyPixels may not work correctly when using a
tilesort configuration.  Since the mural size usually doesn't match
the original window size, there may be any number of inconsistences
between what's drawn with glDrawPixels and what's seen on-screen,
for example.
</p>
<p>
Similarly, glReadPixels may not return the expected results.  For
example, if the application window is 600x400 and you try to read
back the entire 600x400 window, but you're using a tilesort config
with a mural size of 2000x1500, you won't get back the whole image.
</p>


<h2 class="PageSubHeading">Conformance Test Results</h2>

<p>
Here are the results of running the OpenGL conformance tests with
the tilesort SPU (1 tile, 100x100 pixels) with comments explaining
some of the failures:
</p>
<pre>
    Viewport Clamp test (Test number #2) failed.
       Chromium extends the default viewport.
    DrawPixels test (Test number #29) failed.
       GL_DEPTH_COMPONENT test fails because of glRasterPos problem.
    CopyPixels test (Test number #30) failed.
       GL_DEPTH test fails because of glRasterPos problem.
    Bitmap Rasterization test (Test number #31) failed.
       Bitmap renders incorrectly due to glRasterPos problem.
    Display Lists test (Test number #72) failed.
       Known display list problem.
    Texture Edge Clamp test (Test number #75) not supported (not an error).
    Packed Pixels test (Test number #76) failed.
    Texture LOD test (Test number #77) failed.
    Color Table test (Test number #79) not supported (not an error).
    Convolution test (Test number #80) not supported (not an error).
    Convolution Border test (Test number #81) not supported (not an error).
    Histogram test (Test number #82) not supported (not an error).
    MinMax test (Test number #83) not supported (not an error).
    Texture Combine test (Test number #85) not supported (not an error).
    Texture Cube Map test (Test number #86) failed.
    Texture Add test (Test number #88) not supported (not an error).
</pre>
<p>
To show the differences of the tilesort SPU failures, here are the
results of running the OpenGL conformance tests with the tilesort SPU
(2 tiles, 50x100 pixels each) again with comments explaining some
of the failures:
</p>
<pre>
    Viewport Clamp test (Test number #2) failed.
       Chromium extends the default viewport.
    Viewport Transformation test (Test number #8) failed.
       This may never be fixed because of the way the tilesort SPU
       manipulates the OpenGL viewport.
    Buffer Clear test (Test number #9) failed.
    Buffer Invariance test (Test number #14) failed.
       This test happens to draw a point which exactly falls in the
       center of the mural and winds up getting culled.  If the point is
       moved, the test passes.
    Scissor test (Test number #18) failed.
    Stencil Plane Operation test (Test number #22) failed.
    Blend test (Test number #26) failed.
    DrawPixels test (Test number #29) failed.
       (as per single tile above)
    CopyPixels test (Test number #30) failed.
       (as per single tile above)
    Bitmap Rasterization test (Test number #31) failed.
       (as per single tile above)
    Anti-aliased Point test (Test number #33) failed.
    Line Rasterization test (Test number #34) failed.
    Line Stipple test (Test number #35) failed.
       Line stipple can't be properly handled when lines cross tiles.
    Anti-aliased Line test (Test number #36) failed.
    Horizontal and Vertical Line test (Test number #37) failed.
    Triangle Rasterization test (Test number #38) failed.
    Triangle Tile test (Test number #39) failed.
    Triangle Stipple test (Test number #40) failed.
       Polygon stipples must be rotated if tile sizes aren't multiples of 32.
    Anti-aliased Triangles test (Test number #41) failed.
    Quad Rasterization test (Test number #42) failed.
    Polygon Face test (Test number #43) failed.
    Polygon Cull test (Test number #44) failed.
    Polygon Stipple test (Test number #45) failed.
       Polygon stipples must be rotated if tile sizes aren't multiples of 32.
    Polygon Edge test (Test number #46) failed.
    Texture Decal test (Test number #68) failed.
    Texture Border test (Test number #69) failed.
    Mipmaps Selection test (Test number #70) failed.
    Display Lists test (Test number #72) failed.
       Known display list problem.
    Texture Edge Clamp test (Test number #75) failed.
    Packed Pixels test (Test number #76) failed.
    Texture LOD test (Test number #77) failed.
    Rescale Normal test (Test number #78) failed.
    Color Table test (Test number #79) not supported (not an error).
    Convolution test (Test number #80) not supported (not an error).
    Convolution Border test (Test number #81) not supported (not an error).
    Histogram test (Test number #82) not supported (not an error).
    MinMax test (Test number #83) not supported (not an error).
    MultiTexture test (Test number #84) failed.
    Texture Combine test (Test number #85) not supported (not an error).
    Texture Cube Map test (Test number #86) failed.
    Texture Add test (Test number #88) not supported (not an error).
</pre>


<h2 class="PageSubHeading">GLX</h2>

<p>
The following GLX functions are not implemented by Chromium.  Calling
them will result in an error message.  Zero will be returned if the
function returns a value.
</p>

<ul>
<li><code>glXCopyContext</code>
<li><code>glXCreateGLXPixmap</code>
<li><code>glXDestroyGLXPixmap</code>
<li><code>glXWaitGL</code>
<li><code>glXWaitX</code>
</ul>

<p>
Additionally, none of the GLX 1.3 functions related to Pbuffers or
visual configs are implemented.
</p>

<p>
GLX visuals present a problem for Chromium.  The basic problem is that
the visuals available on the application host display may not match
the visuals of the rendering host display(s).  If the application
queries the GLX visual attributes of the local host display it may
find a visual that has no equal on the rendering hosts.  It may even
be the case that the local host display has no GLX capabilities at
all.  In a heterogeneous cluster, the rendering hosts themselves may
have different visual capabilities.
</p>
<p>
Ideally, when the application queries GLX visuals and their attributes
it should get results which reflect the capabilities of the rendering
host displays.  The results should be the minimums of the union of the
rendering hosts.  For example, if there are four rendering hosts and
three of them support accumulation buffers but one doesn't then the
application should be told that no accumulation buffer exists.
</p>
<p>
This level of sophistication does not yet exist in the Chromium GLX
code.  Therefore, it's possible for an application to determine that
there are visual capabilities that don't really exist, causing the
application to eventually abort.  Luckily, all modern graphics
hardware supports stencil buffers, alpha channels and accumulation
buffers so this is seldom a problem.  Trying to use stereo and
multisample color buffers may be problematic, however.
</p>


<h2 class="PageSubHeading">WGL</h2>

<p>
The following WGL functions are not implemented by Chromium.  Calling
them will result in an error message.  Zero will be returned if the
function returns a value.
</p>

<ul>
<li><code>wglShareLists</code>
<li><code>wglCopyContext</code>
<li><code>wglCreateLayerContext</code>
<li><code>wglUseFontBitmapsA</code>
<li><code>wglUseFontBitmapsW</code>
<li><code>wglDescribeLayerPlane</code>
<li><code>wglSetLayerPaletteEntries</code>
<li><code>wglGetLayerPaletteEntries</code>
<li><code>wglRealizeLayerPalette</code>
<li><code>wglSwapMultipleBuffers</code>
<li><code>wglUseFontOutlinesA</code>
<li><code>wglUseFontOutlinesW</code>
<li><code>wglSwapLayerBuffers</code>
</ul>


</body>

</html>