1.0-8756 GLSL internal compiler errors

tachyon_john · 05-18-06, 03:39 PM

Hi,
The 8756 drivers fail to compile a shader, printing a large infolog with "internal compiler error" messages. The same shader compiled correctly with older driver versions, and with other video cards. The nvidia-bug-report.log file is too large for the forum's upload limit, so I've put it in zip/tar files along with test data to reproduce the problem, and the shader compiler error messages for you. The files are posted for you here, and include
instructions:
http://www.ks.uiuc.edu/Research/vmd/alpha/test/

I've attached the shader compiler errors as an attachment since they are small.

John Stone

netllama · 05-18-06, 04:57 PM

John,
I'm attempting to reproduce this, however I'm not that familiar with VMD. When I go to File -> Load State and select your testglsl.vmd file all I'm getting are errors complaining that the file output.psf could not be read. Please detail how I can reproudce the problem you're reporting.

For future reference, you can zip up your bug report to attach it here.

Also, is this problem different than what you've posted in this thread?
http://www.nvnews.net/vbulletin/showthread.php?t=70279

Thanks,
Lonni

tachyon_john · 05-18-06, 05:03 PM

Lonni,
Follow the instructions in the README.TXT I included... Don't use the GUI, use
the VMD text console or the command line (you have to be in the unpacked
testglsl directory before loading the saved state..)

This is a new different shader compilation problem specific to 8756, different
from the one I reported before. I'm not using any weird locale, so I'm not sure what the problem is.

This test dataset will reproduce both the older bugs I reported previously,
as well as these new ones (same shader source code). The test data
is pretty thorough in exercising most of the shader code. (you could turn
fog on/off manually to test the remaining bits...)

John

tachyon_john · 05-18-06, 05:04 PM

Lonni,
I forgot to mention, I tried uploading the test data as a ZIP file, but the file size limit on this forum is only 100KB, and the test dataset is several megabytes even compressed, so that wasn't an option.

tachyon_john · 05-18-06, 06:20 PM

>
> Also, is this problem different than what you've posted in this thread?
> http://www.nvnews.net/vbulletin/showthread.php?t=70279
>

The difference between these two bug reports is that this one is regarding the shader compilation failure for the vmdsphere.frag shader (see the shaders subdirectory
I included in the example data).

The other bug refers to behavior associated with texturing with the vmd.frag
shader which is use for all non-sphere geometry. (if the sphere shader compiled, I expect it would exhibit the same texturing problem as well however)

John

netllama · 05-19-06, 11:54 AM

John,
Thanks for clarifying. I was able to reproduce both of these problems with 1.0-8756, however both no longer reproduced using the latest internal development driver. The next released driver will, unfortunately, not include this fix, however the driver after that should.

Thanks,
Lonni

mooninite · 05-19-06, 03:11 PM

Quote:

Originally Posted by netllama

The next released driver will, unfortunately, not include this fix, however the driver after that should.

I hope it doesn't take until Christmas to get out that driver.

tachyon_john · 05-19-06, 05:31 PM

Quote:

Originally Posted by netllama

John,
Thanks for clarifying. I was able to reproduce both of these problems with 1.0-8756, however both no longer reproduced using the latest internal development driver. The next released driver will, unfortunately, not include this fix, however the driver after that should.

Thanks,
Lonni

Lonni,
Do these upcoming driver versions have version numbers assigned to them yet? It would be really nice if I knew which version number (even if it's coming up in a month or two) will work correctly so I can tell VMD user's what version to wait for when the subject comes up.

John

netllama · 05-19-06, 05:35 PM

The driver that could potentially include this fix would be a 1.0-9xxx series driver.

Thanks,
Lonni

maibaum · 07-17-06, 08:27 PM

Is there any news as to when this new driver will be available? I am trying to get vmd to run with the 1.0-8762 driver on x86_64, but it fails for this reason...

05-18-06, 04:57 PM	#2
netllama NVIDIA Corporation Join Date: Dec 2004 Posts: 8,763	Re: 1.0-8756 GLSL internal compiler errors John, I'm attempting to reproduce this, however I'm not that familiar with VMD. When I go to File -> Load State and select your testglsl.vmd file all I'm getting are errors complaining that the file output.psf could not be read. Please detail how I can reproudce the problem you're reporting. For future reference, you can zip up your bug report to attach it here. Also, is this problem different than what you've posted in this thread? http://www.nvnews.net/vbulletin/showthread.php?t=70279 Thanks, Lonni

05-18-06, 05:03 PM	#3
tachyon_john Registered User Join Date: Jan 2006 Posts: 32	Re: 1.0-8756 GLSL internal compiler errors Lonni, Follow the instructions in the README.TXT I included... Don't use the GUI, use the VMD text console or the command line (you have to be in the unpacked testglsl directory before loading the saved state..) This is a new different shader compilation problem specific to 8756, different from the one I reported before. I'm not using any weird locale, so I'm not sure what the problem is. This test dataset will reproduce both the older bugs I reported previously, as well as these new ones (same shader source code). The test data is pretty thorough in exercising most of the shader code. (you could turn fog on/off manually to test the remaining bits...) John

05-18-06, 05:04 PM	#4
tachyon_john Registered User Join Date: Jan 2006 Posts: 32	Re: 1.0-8756 GLSL internal compiler errors Lonni, I forgot to mention, I tried uploading the test data as a ZIP file, but the file size limit on this forum is only 100KB, and the test dataset is several megabytes even compressed, so that wasn't an option.

05-18-06, 06:20 PM	#5
tachyon_john Registered User Join Date: Jan 2006 Posts: 32	Re: 1.0-8756 GLSL internal compiler errors > > Also, is this problem different than what you've posted in this thread? > http://www.nvnews.net/vbulletin/showthread.php?t=70279 > The difference between these two bug reports is that this one is regarding the shader compilation failure for the vmdsphere.frag shader (see the shaders subdirectory I included in the example data). The other bug refers to behavior associated with texturing with the vmd.frag shader which is use for all non-sphere geometry. (if the sphere shader compiled, I expect it would exhibit the same texturing problem as well however) John

05-19-06, 11:54 AM	#6
netllama NVIDIA Corporation Join Date: Dec 2004 Posts: 8,763	Re: 1.0-8756 GLSL internal compiler errors John, Thanks for clarifying. I was able to reproduce both of these problems with 1.0-8756, however both no longer reproduced using the latest internal development driver. The next released driver will, unfortunately, not include this fix, however the driver after that should. Thanks, Lonni

05-19-06, 05:35 PM	#9
netllama NVIDIA Corporation Join Date: Dec 2004 Posts: 8,763	Re: 1.0-8756 GLSL internal compiler errors The driver that could potentially include this fix would be a 1.0-9xxx series driver. Thanks, Lonni

07-17-06, 08:27 PM	#10
maibaum Registered User Join Date: Feb 2004 Posts: 3	Re: 1.0-8756 GLSL internal compiler errors Is there any news as to when this new driver will be available? I am trying to get vmd to run with the 1.0-8762 driver on x86_64, but it fails for this reason...