Hello Expert
LINUX
gpu version is 1.13.5776728
SDK 7.2
The problem is as shown below. Currently, there is a very low probability that the problem will occur
7416.dmesg.logpvrlogdump_2203011033.txt.gz
This thread has been locked.
If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.
Hello Expert
LINUX
gpu version is 1.13.5776728
SDK 7.2
The problem is as shown below. Currently, there is a very low probability that the problem will occur
7416.dmesg.logpvrlogdump_2203011033.txt.gz
Hello Li,
This is a new issue I have not seen before. Usually GPU issues are followed by PVR logs in the console. One good way to check this is to capture the input and output of the GPU to see if the scene came in distorted to the GPU or if the GPU is the culprit creating the bad output.
Do you have any details on how many units are showing this behavior?
Regards,
Erick
Hello, Eric,
We obtained the image before the GPU input,image not distorted。
There is no PVR related printing in stdout. the probability of occurrence in the product is very low. Less than ten motherboards have this problem.We would like to prioritize whether there is a hardware issue
Details:Between one to two hours of normal operation of the board, the appearance of the above image suddenly appears. During this period, you can switch views or execute 3D animations. it will automatically return to its normal display at random time.
There are logs we have collected on it. Do you need any other content?
Li,
Would you be able to capture a PVRCarbon of this activity? Instructions are here: https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1218307/faq-how-can-i-analyze-a-gpu-graphics-process-unit-driver-or-hardware-issue
Last section describes it.
It will replicate the OpenGL calls, so we can see if the GPU is the culprit. If this does not work due to your implementation, we can look at alternatives.
Regards,
Erick
Erick,
I have obtained the information using the tool. Please help me check it,The file is linked below,thanks.
Link:pan.baidu.com/.../1piCA4sBlNiCRxegeCAAtVw
Extracted code:ykzr
Li,
I'm not able to access the image. Could you please upload it to the E2E ticket? It's available for upload on the Insert->Image/Video/file option
Then you click on the Upload button:
Thanks,
Erick
Li,
Thanks. I'm getting errors reading back the contents of the file, but I believe it is a limitation on the tool itself. Let me check with our vendor and get back to you soon. And I'm checking if there are any other traces that might be useful.
Thanks,
Erick
Li,
The traces that would be useful would be if the GPU reports an error, you'll see the debug output in the console.
But if this issue does not report an error, it indicates something else could be the issue. Do you know how reproduceable the error is? We will need to catch the error. If you have how long it usually takes to replicate, and how long it replicates, this would be valuable information.
Thanks,
Erick
Li,
The logs would occur on the console. But if they don't then the GPU driver does not indicate that it has hit an error.
Is this occurring on all platforms? Or just one?
Regards,
Erick
Li,
Interesting, calling glFinish() twice should not do anything, but I can check with our GPU vendor on this point if that should have any special effects.
Thanks,
Erick
Li,
Thanks for the logs, I'll get these analyzed and provide further instruction.
Regards,
Erick
Hi, Eric,
Today's test, there is a flower screen problem.The patch package added for this test. The name of the patch package is patch_11.13_PP131820_CL633615.tgz.
Help analyze the problem. thank you.
Flower screen front :HWR Event Count:1
Flower Screen :HWR Event Count:2
Screen restoration : HWR Event Count:3
Hello Li,
The patch package added for this test. The name of the patch package is patch_11.13_PP131820_CL633615.tgz.
Where did you get this patch? Are these pre-built GPU libraries provided by someone?
Regards,
Erick
Hi, Ercik
This the UM with patch for Baidu previous black display issue. Patch is PP131820_CL6336153_reduce_tiles_in_flight.patch. This patch fix the black display issue.
Link is below:
Regards
Zekun
Li, Zekun,
Understood.
Can you please run the following command before you start your application:
echo Y > /sys/kernel/debug/pvr/apphint/0/AssertOnHWRTrigger
pvrdebug -loggroups main,mts,cleanup,csw,bif,pm,rtd,spm,pow,hwr
Then, collect the pvrlogdump when you experience a hardware recovery. The first command will cause the GPU to freeze and save the firmware state when an issue occurs. That will provide us a snapshot and give us something to analyze.
Thanks,
Erick
Hi,Ercik
In the process of repetition avm does not have a Mosaic screen, but the screen continues to be black once. I have obtained the relevant log according to the method, please help analyze it.
Thanks.
Li,
Thanks, I'm sharing these logs with the team for analysis. The black screen is expected as the FW is frozen using this method and no more GPU processing is handled.
Regards,
Erick
Biao,
The logs were just sent yesterday. Please expect delays as the team in the US and in Europe that are looking into this might be on holidays.
Thanks,
Erick
Hello Biao,
This thread got locked, and the team has released some new updates to the GPU driver that might help this situation.
Is this still an issue?
Thanks,
Erick
Biao,
Closing this thread, please re-open or submit a new thread if you still need assistance.
Thanks,
Erick