GPU Assertion error?

[chrono-cmu]

Hello, I am running a modified version of the cone penetrometer demo code, and when I attempt to run my code (no changes from last time I successfully ran it) I get the following error:

terminate called after throwing an instance of 'std::runtime_error'
  what():  GPU Assertion: forward compatibility was attempted on non supported HW. This happened in /home/sidney/DEM-Engine/src/core/utils/GpuManager.cpp:46

You can check out the troubleshoot section of DEME to see if it helps, or post this error on Chrono's forum https://groups.google.com/g/projectchrono

The mentioned file path does not exist on my computer, leaving me a little stumped. Any help would be greatly appreciated. Thanks.

[Ruochun Zhang]

This appears to be related to some environmental changes as it failed at the first CUDA call. Did you compile the C++ version of the code or are you using PyDEME? And if the former, it's weird since that file is a part of DEME and you should have it, unless, for example, you compiled the code last time in a container and you did not mount the appropriate drive this time.

If you are using the C++ version, then I assume you are using a cluster. You should make sure you are loading the same CUDA environment as last time (assuming they are managed by something like modules). Also try requesting the same GPU as last time (this is not required by DEME, but it could help us rule out some possibilities). If you are not sure about how you did it last time, it's always viable to delete everything in the build directory and try rebuilding it with the environment you have now; in which case, showing us the cmake options you used can help, too.

If you are using PyDEME, then all the above still apply, and you should remember to load the same conda environment you did last time. Again, if not sure how you did last time, you can create a new conda env to see if you can reinstall it and then reproduce this problem.

Thank you,

Ruochun

[chrono-cmu]

Awesome, thanks for the help!