This page lists known issues with SPEC Viewperf 11 when running on Mesa-based drivers.
The Viewperf data sets are basically GL API traces that are recorded from CAD applications, then replayed in the Viewperf framework.
The primary problem with these traces is they blindly use features and OpenGL extensions that were supported by the OpenGL driver when the trace was recorded, but there's no checks to see if those features are supported by the driver when playing back the traces with Viewperf.
These issues have been reported to the SPEC organization in the hope that they'll be fixed in the future.
Some of the Viewperf tests use a lot of memory. At least 2GB of RAM is recommended.
Catia-03 test 2
This test creates over 38000 vertex buffer objects. On some systems this can exceed the maximum number of buffer allocations. Mesa generates GL_OUT_OF_MEMORY errors in this situation, but Viewperf does no error checking and continues. When this happens, some drawing commands become no-ops. This can also eventually lead to a segfault either in Viewperf or the Mesa driver.
Catia-03 tests 3, 4, 8
When Mesa tries to compile the vertex/fragment programs it generates errors (which Viewperf ignores). Subsequent drawing calls become no-ops and the rendering is incorrect.
sw-02 tests 1, 2, 4, 6
These tests depend on the GL_NV_primitive_restart extension.
If the Mesa driver doesn't support this extension the rendering will be incorrect and the test will fail.
Also, the color of the line drawings in test 2 seem to appear in a random color. This is probably due to some uninitialized state somewhere.
sw-02 test 6
The lines drawn in this test appear in a random color. That's because texture mapping is enabled when the lines are drawn, but no texture image is defined (glTexImage2D() is called with pixels=NULL). Since GL says the contents of the texture image are undefined in that situation, we get a random color.
Lightwave-01 test 3
This test uses a number of mipmapped textures, but the textures are incomplete because the last/smallest mipmap level (1 x 1 pixel) is never specified.
A trace captured with API trace shows this sequences of calls like this:
2504 glBindTexture(target = GL_TEXTURE_2D, texture = 55) 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)) 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)) 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)) [...] 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)) 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)) 2514 glTexParameteri(target = GL_TEXTURE_2D, pname = GL_TEXTURE_MIN_FILTER, param = GL_LINEAR_MIPMAP_LINEAR) 2515 glTexParameteri(target = GL_TEXTURE_2D, pname = GL_TEXTURE_WRAP_S, param = GL_REPEAT) 2516 glTexParameteri(target = GL_TEXTURE_2D, pname = GL_TEXTURE_WRAP_T, param = GL_REPEAT) 2517 glTexParameteri(target = GL_TEXTURE_2D, pname = GL_TEXTURE_MAG_FILTER, param = GL_NEAREST)
Note that one would expect call 2514 to be glTexImage(level=9, width=1, height=1) but it's not there.
The minification filter is GL_LINEAR_MIPMAP_LINEAR and the texture's GL_TEXTURE_MAX_LEVEL is 1000 (the default) so a full mipmap is expected.
Later, these incomplete textures are bound before drawing calls. According to the GL specification, if a fragment program or fragment shader is being used, the sampler should return (0,0,0,1) ("black") when sampling from an incomplete texture. This is what Mesa does and the resulting rendering is darker than it should be.
It appears that NVIDIA's driver (and possibly AMD's driver) detects this case and returns (1,1,1,1) (white) which causes the rendering to appear brighter and match the reference image (however, AMD's rendering is much brighter than NVIDIA's).
If the fallback texture created in _mesa_get_fallback_texture() is initialized to be full white instead of full black the rendering appears correct. However, we have no plans to implement this work-around in Mesa.
Maya-03 test 2
This test makes some unusual calls to glRotate. For example:
glRotate(50, 50, 50, 1); glRotate(100, 100, 100, 1); glRotate(52, 52, 52, 1);
These unusual values lead to invalid modelview matrices. For example, the last glRotate command above produces this matrix with Mesa:
1.08536e+24 2.55321e-23 -0.000160389 0 5.96937e-25 1.08536e+24 103408 0 103408 -0.000160389 1.74755e+09 0 0 0 0 nanand with NVIDIA's OpenGL:
1.4013e-45 0 -nan 0 0 1.4013e-45 1.4013e-45 0 1.4013e-45 -nan 1.4013e-45 0 0 0 0 1.4013e-45
This causes the object in question to be drawn in a strange orientation and with a semi-random color (between white and black) since GL_FOG is enabled.
Proe-05 test 1
This uses depth testing but there's two problems:
- The glXChooseFBConfig() call doesn't request a depth buffer
- The test never calls glClear(GL_DEPTH_BUFFER_BIT) to initialize the depth buffer
If the chosen visual does not have a depth buffer, you'll see the wireframe car model but it won't be rendered correctly.If (by luck) the chosen visual has a depth buffer, its initial contents will be undefined so you may or may not see parts of the model.
Interestingly, with NVIDIA's driver most visuals happen to have a depth buffer and apparently the contents are initialized to 1.0 by default so this test just happens to work with their drivers.
Finally, even if a depth buffer was requested and the glClear(GL_COLOR_BUFFER_BIT) calls were changed to glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT) the problem still wouldn't be fixed because GL_DEPTH_WRITEMASK=GL_FALSE when glClear is called so clearing the depth buffer would be a no-op anyway.