• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1Viewperf Issues
2===============
3
4This page lists known issues with `SPEC Viewperf
511 <https://www.spec.org/gwpg/gpc.static/vp11info.html>`__ and `SPEC
6Viewperf 12 <https://www.spec.org/gwpg/gpc.static/vp12info.html>`__ when
7running on Mesa-based drivers.
8
9The Viewperf data sets are basically GL API traces that are recorded
10from CAD applications, then replayed in the Viewperf framework.
11
12The primary problem with these traces is they blindly use features and
13OpenGL extensions that were supported by the OpenGL driver when the
14trace was recorded, but there's no checks to see if those features are
15supported by the driver when playing back the traces with Viewperf.
16
17These issues have been reported to the SPEC organization in the hope
18that they'll be fixed in the future.
19
20Viewperf 11
21-----------
22
23Some of the Viewperf 11 tests use a lot of memory. At least 2GB of RAM
24is recommended.
25
26Catia-03 test 2
27~~~~~~~~~~~~~~~
28
29This test creates over 38000 vertex buffer objects. On some systems this
30can exceed the maximum number of buffer allocations. Mesa generates
31GL_OUT_OF_MEMORY errors in this situation, but Viewperf does no error
32checking and continues. When this happens, some drawing commands become
33no-ops. This can also eventually lead to a segfault either in Viewperf
34or the Mesa driver.
35
36Catia-03 tests 3, 4, 8
37~~~~~~~~~~~~~~~~~~~~~~
38
39These tests use features of the
40`GL_NV_fragment_program2 <https://www.opengl.org/registry/specs/NV/fragment_program2.txt>`__
41and
42`GL_NV_vertex_program3 <https://www.opengl.org/registry/specs/NV/vertex_program3.txt>`__
43extensions without checking if the driver supports them.
44
45When Mesa tries to compile the vertex/fragment programs it generates
46errors (which Viewperf ignores). Subsequent drawing calls become no-ops
47and the rendering is incorrect.
48
49sw-02 tests 1, 2, 4, 6
50~~~~~~~~~~~~~~~~~~~~~~
51
52These tests depend on the
53`GL_NV_primitive_restart <https://www.opengl.org/registry/specs/NV/primitive_restart.txt>`__
54extension.
55
56If the Mesa driver doesn't support this extension the rendering will be
57incorrect and the test will fail.
58
59Also, the color of the line drawings in test 2 seem to appear in a
60random color. This is probably due to some uninitialized state
61somewhere.
62
63sw-02 test 6
64~~~~~~~~~~~~
65
66The lines drawn in this test appear in a random color. That's because
67texture mapping is enabled when the lines are drawn, but no texture
68image is defined (glTexImage2D() is called with pixels=NULL). Since GL
69says the contents of the texture image are undefined in that situation,
70we get a random color.
71
72Lightwave-01 test 3
73~~~~~~~~~~~~~~~~~~~
74
75This test uses a number of mipmapped textures, but the textures are
76incomplete because the last/smallest mipmap level (1 x 1 pixel) is never
77specified.
78
79A trace captured with `API
80trace <https://github.com/apitrace/apitrace>`__ shows this sequences of
81calls like this:
82
83::
84
85   2504 glBindTexture(target = GL_TEXTURE_2D, texture = 55)
86   2505 glTexImage2D(target = GL_TEXTURE_2D, level = 0, internalformat = GL_RGBA, width = 512, height = 512, border = 0, format = GL_RGB, type = GL_UNSIGNED_SHORT, pixels = blob(1572864))
87   2506 glTexImage2D(target = GL_TEXTURE_2D, level = 1, internalformat = GL_RGBA, width = 256, height = 256, border = 0, format = GL_RGB, type = GL_UNSIGNED_SHORT, pixels = blob(393216))
88   2507 glTexImage2D(target = GL_TEXTURE_2D, level = 2, internalformat = GL_RGBA, width = 128, height = 128, border = 0, format = GL_RGB, type = GL_UNSIGNED_SHORT, pixels = blob(98304))
89   [...]
90   2512 glTexImage2D(target = GL_TEXTURE_2D, level = 7, internalformat = GL_RGBA, width = 4, height = 4, border = 0, format = GL_RGB, type = GL_UNSIGNED_SHORT, pixels = blob(96))
91   2513 glTexImage2D(target = GL_TEXTURE_2D, level = 8, internalformat = GL_RGBA, width = 2, height = 2, border = 0, format = GL_RGB, type = GL_UNSIGNED_SHORT, pixels = blob(24))
92   2514 glTexParameteri(target = GL_TEXTURE_2D, pname = GL_TEXTURE_MIN_FILTER, param = GL_LINEAR_MIPMAP_LINEAR)
93   2515 glTexParameteri(target = GL_TEXTURE_2D, pname = GL_TEXTURE_WRAP_S, param = GL_REPEAT)
94   2516 glTexParameteri(target = GL_TEXTURE_2D, pname = GL_TEXTURE_WRAP_T, param = GL_REPEAT)
95   2517 glTexParameteri(target = GL_TEXTURE_2D, pname = GL_TEXTURE_MAG_FILTER, param = GL_NEAREST)
96
97Note that one would expect call 2514 to be glTexImage(level=9, width=1,
98height=1) but it's not there.
99
100The minification filter is GL_LINEAR_MIPMAP_LINEAR and the texture's
101GL_TEXTURE_MAX_LEVEL is 1000 (the default) so a full mipmap is expected.
102
103Later, these incomplete textures are bound before drawing calls.
104According to the GL specification, if a fragment program or fragment
105shader is being used, the sampler should return (0,0,0,1) ("black") when
106sampling from an incomplete texture. This is what Mesa does and the
107resulting rendering is darker than it should be.
108
109It appears that NVIDIA's driver (and possibly AMD's driver) detects this
110case and returns (1,1,1,1) (white) which causes the rendering to appear
111brighter and match the reference image (however, AMD's rendering is
112*much* brighter than NVIDIA's).
113
114If the fallback texture created in \_mesa_get_fallback_texture() is
115initialized to be full white instead of full black the rendering appears
116correct. However, we have no plans to implement this work-around in
117Mesa.
118
119Maya-03 test 2
120~~~~~~~~~~~~~~
121
122This test makes some unusual calls to glRotate. For example:
123
124.. code-block:: c
125
126   glRotate(50, 50, 50, 1);
127   glRotate(100, 100, 100, 1);
128   glRotate(52, 52, 52, 1);
129
130These unusual values lead to invalid modelview matrices. For example,
131the last glRotate command above produces this matrix with Mesa:
132
133.. math::
134
135   \begin{matrix}
136   1.08536 \times 10^{24} & 2.55321 \times 10^{-23} & -0.000160389         & 0\\
137   5.96937 \times 10^{25} & 1.08536 \times 10^{24}  & 103408               & 0\\
138                   103408 & -0.000160389            & 1.74755\times 10^{9} & 0\\
139   0                      &                       0 &                      0 & nan
140   \end{matrix}
141
142and with NVIDIA's OpenGL:
143
144.. math::
145
146   \begin{matrix}
147   1.4013 \times 10^{-45} &                      0 &                   -nan & 0\\
148                        0 & 1.4013 \times 10^{-45} & 1.4013 \times 10^{-45} & 0\\
149   1.4013 \times 10^{-45} &                   -nan & 1.4013 \times 10^{-45} & 0\\
150                        0 &                      0 &                      0 & 1.4013 \times 10^{-45}
151   \end{matrix}
152
153This causes the object in question to be drawn in a strange orientation
154and with a semi-random color (between white and black) since GL_FOG is
155enabled.
156
157Proe-05 test 1
158~~~~~~~~~~~~~~
159
160This uses depth testing but there's two problems:
161
162#. The glXChooseFBConfig() call doesn't request a depth buffer
163#. The test never calls glClear(GL_DEPTH_BUFFER_BIT) to initialize the
164   depth buffer
165
166If the chosen visual does not have a depth buffer, you'll see the
167wireframe car model but it won't be rendered correctly.
168
169If (by luck) the chosen visual has a depth buffer, its initial contents
170will be undefined so you may or may not see parts of the model.
171
172Interestingly, with NVIDIA's driver most visuals happen to have a depth
173buffer and apparently the contents are initialized to 1.0 by default so
174this test just happens to work with their drivers.
175
176Finally, even if a depth buffer was requested and the
177glClear(GL_COLOR_BUFFER_BIT) calls were changed to
178glClear(GL_COLOR_BUFFER_BIT \| GL_DEPTH_BUFFER_BIT) the problem still
179wouldn't be fixed because GL_DEPTH_WRITEMASK=GL_FALSE when glClear is
180called so clearing the depth buffer would be a no-op anyway.
181
182Proe-05 test 6
183~~~~~~~~~~~~~~
184
185This test draws an engine model with a two-pass algorithm. The first
186pass is drawn with polygon stipple enabled. The second pass is drawn
187without polygon stipple but with blending and GL_DEPTH_FUNC=GL_LEQUAL.
188If either of the two passes happen to use a software fallback of some
189sort, the Z values of fragments may be different between the two passes.
190This leads to incorrect rendering.
191
192For example, the VMware SVGA Gallium driver uses a special semi-fallback
193path for drawing with polygon stipple. Since the two passes are rendered
194with different vertex transformation implementations, the rendering
195doesn't appear as expected. Setting the SVGA_FORCE_SWTNL environment
196variable to 1 will force the driver to use the software vertex path all
197the time and clears up this issue.
198
199According to the OpenGL invariance rules, there's no guarantee that the
200pixels produced by these two rendering states will match. To achieve
201invariance, both passes should enable polygon stipple and blending with
202appropriate patterns/modes to ensure the same fragments are produced in
203both passes.
204
205Viewperf 12
206-----------
207
208Note that Viewperf 12 only runs on 64-bit Windows 7 or later.
209
210catia-04
211~~~~~~~~
212
213One of the catia tests calls wglGetProcAddress() to get some
214GL_EXT_direct_state_access functions (such as glBindMultiTextureEXT) and
215some GL_NV_half_float functions (such as glMultiTexCoord3hNV). If the
216extension/function is not supported, wglGetProcAddress() can return
217NULL. Unfortunately, Viewperf doesn't check for null pointers and
218crashes when it later tries to use the pointer.
219
220Another catia test uses OpenGL 3.1's primitive restart feature. But when
221Viewperf creates an OpenGL context, it doesn't request version 3.1 If
222the driver returns version 3.0 or earlier all the calls related to
223primitive restart generate an OpenGL error. Some of the rendering is
224then incorrect.
225
226energy-01
227~~~~~~~~~
228
229This test creates a 3D luminance texture of size 1K x 1K x 1K. If the
230OpenGL driver/device doesn't support a texture of this size the
231glTexImage3D() call will fail with GL_INVALID_VALUE or GL_OUT_OF_MEMORY
232and all that's rendered is plain white polygons. Ideally, the test would
233use a proxy texture to determine the max 3D texture size. But it does
234not do that.
235
236maya-04
237~~~~~~~
238
239This test generates many GL_INVALID_OPERATION errors in its calls to
240glUniform(). Causes include:
241
242-  Trying to set float uniforms with glUniformi()
243-  Trying to set float uniforms with glUniform3f()
244-  Trying to set matrix uniforms with glUniform() instead of
245   glUniformMatrix().
246
247Apparently, the indexes returned by glGetUniformLocation() were
248hard-coded into the application trace when it was created. Since
249different implementations of glGetUniformLocation() may return different
250values for any given uniform name, subsequent calls to glUniform() will
251be invalid since they refer to the wrong uniform variables. This causes
252many OpenGL errors and leads to incorrect rendering.
253
254medical-01
255~~~~~~~~~~
256
257This test uses a single GLSL fragment shader which contains a GLSL 1.20
258array initializer statement, but it neglects to specify ``#version 120``
259at the top of the shader code. So, the shader does not compile and all
260that's rendered is plain white polygons.
261
262Also, the test tries to create a very large 3D texture that may exceed
263the device driver's limit. When this happens, the glTexImage3D call
264fails and all that's rendered is a white box.
265
266showcase-01
267~~~~~~~~~~~
268
269This is actually a DX11 test based on Autodesk's Showcase product. As
270such, it won't run with Mesa.
271