DEM-Engine Issues

[Mahmoud Yaseen]

Hello All,

I am trying to run the tutorial BallDrop2D on the github repo, When I try to compile the .cpp file, I get an error stating that cmake version is not compatible. Needs minimum version 3.27. The version I have is up to date 3.27.7. 

Also, I tried to run it PyDEME version, my GPU is recongized and my cuda version is 12.2. I tried older versions like 11, but when I run the example I get runtime error and cuda version is defined. My question is what is the best cuda version suitable for PyDEME that will not cause problems?

Thanks

[Ruochun Zhang]

Hi Mahmoud,

With what we have here, I feel both issues are related to multiple installations.

For cmake, it could be that you have multiple cmake installations, in which case you should specify the one you want to use in PATH. Or if you are using ccmake, perhaps it's linked against a cmake backend which is old. Reinstalling them should always work... The README document of DEME has a link to a guide on updating cmake to the newest. It's written for WSL, but works for Linux as well.

For CUDA, I'm pretty sure 11.3 and 12.1~12.3 do not work, which may include the versions you tried. This is a limitation I cannot overcome and you just have to use a different version. I would just recommend the newest CUDA; or try 12.0, if you want to be safe. The most important thing is to set the CUDA-related environment variables correctly before compilation, if you have multiple installations.  The README document of DEME also contains some guidance on how to do that.

Ruochun

[Kelvin Ding]

Hi Ruochun, I encountered a similar issue when compiling the DEM-engine. My CUDA version is 13.0. After generating the .sln file via CMake, I tried to build the solution in MSVS 2022, but the build failed with the error message shown in the image below. How should I resolve this? Best regards, Ding

[Ruochun Zhang]

Hi Ding,

CUDA13 removed a bunch of things DEME currently uses. We should be able to push a new version in a few days which fixes this problem and provides huge performance boosts. I'll reply here when it is out. So I'd say you either wait a few days, or use CUDA 12.8 or below.

Thank you,

Ruochun