Hi Sarina, If I remember correctly, CUDA11.2 has known issues with quite some software packages, including Chrono::GPU. CUDA11.8 should work well. But if it's not creating problems for you, that's probably fine.
Typical problems with meshes that can affect DEM simulations might include inverted face normals (note the normals are determined by the right-hand rule, not the nodal normals in obj files, because that's for rendering), extremely spiky surfaces, double/invisible layers of surfaces, not being watertight, etc. You know I am a Rhino user. For me, usually using *dir* command in Rhino to check the face normals, then trying to use *QuadRemesh* to see if the mesh can be successfully reconstructed would reveal and provide fixes for most mesh problems. If you use other tools the process may be similar. However, I don't think the texture is a problem since Chrono::GPU won't read it anyway, and if it indeed affects, you should be able to see that in a rendering of the mesh. A side note: I forgot how well C::GPU loads the mesh if there are multiple meshes in a mesh file, but you can make sure there is only one mesh in the file to rule out this possibility. Thank you, Ruochun On Friday, October 20, 2023 at 7:55:08 PM UTC-5 [email protected] wrote: > Hi Ruochun, > Both times with 11. previously it was 11.2, but now it is 11.8. > Took me a while, but I figured the problem was the mesh ( obj file), and > the code ran just fine with the previous mesh file. I have the exact same > design but used a different software! > I am trying to figure out that maybe the texture is being restored in obj > somehow. > Let me know if you have any ideas on this! > > > On Tuesday, October 17, 2023 at 9:55:32 PM UTC-7 Ruochun Zhang wrote: > >> Hi Sarina, >> >> If you are sure that the same script runs differently on V100 and A100, >> then it could be related to the CUDA version. Did you compile the code with >> CUDA 11 or above? >> >> Other than this, I don't have a clear idea. If the script can be shared, >> we might understand it better. >> >> Thank you, >> Ruochun >> >> On Tuesday, October 17, 2023 at 7:16:30 PM UTC-5 [email protected] >> wrote: >> >>> Correction: >>> BallCosim.cpp, not the balldrop.cpp >>> >>> On Tuesday, October 17, 2023 at 5:13:16 PM UTC-7 sarina shahhosseini >>> wrote: >>> >>>> Hi all, >>>> I've been working on a simulation project involving a system of >>>> connected rigid bodies in ::GPU (using ChLinkMotorRotation and >>>> ChLinkMotorLinear). The code structure is similar to the balldrop.cpp >>>> demo, >>>> with the difference being that it involves a connected system of rigid >>>> bodies controlled by motors. >>>> I've recently encountered the following error during the simulation: >>>> >>>> *ERROR! Sphere 440770 has invalid SD 4294967295, max is 139920.* >>>> >>>> This error occurs approximately 1 second into the simulation when the >>>> robot is already in contact with particles ( So the particle >>>> initialization >>>> went smoothly, and the rendered output seems reasonable). So, I am unsure >>>> why the particle is in an invalid subdomain. I've tried reducing the time >>>> step to a significantly smaller value (1e-7) and simplifying the mesh, but >>>> the issue persists. >>>> I previously used a V100 GPU without encountering this issue, but now >>>> I've switched to an A100 GPU. Could this somehow be related? >>>> >>>> I would appreciate your comments! >>>> >>> -- You received this message because you are subscribed to the Google Groups "ProjectChrono" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/projectchrono/61297268-c00f-4f90-b9e0-2ce6ef2f438an%40googlegroups.com.
