CHARMM-GUI Issue

[Kyle Kihn]

Hi all,
I am having when attempting to run outputs from the CHARMM-GUI (generated for openMM). I can get a simulation to run on the GPU if I create my own input files. my nvcc --version is 10.0.130 and the pythom -m simtktestInstallation is normal. However when using the CHARMM-GUI I get the following error.
Traceback (most recent call last):
File "openmm_run.py", line 74, in <module>
simulation = Simulation(psf.topology, system, integrator, platform, prop)
File "/home/dderedge/Desktop/anaconda3/envs/charmmMD/lib/python3.7/site-packages/simtk/openmm/app/simulation.py", line 105, in __init__
self.context = mm.Context(self.system, self.integrator, platform, platformProperties)
File "/home/dderedge/Desktop/anaconda3/envs/charmmMD/lib/python3.7/site-packages/simtk/openmm/openmm.py", line 18608, in __init__
this = _openmm.new_Context(*args)s
Exception: Error launching CUDA compiler: 256
/tmp/openmmTempKernel0x55cc511f2f30_7018.cu(1002): error: calling a constexpr __host__ function("fmin") from a __global__ function("computeBondedForces") is not allowed. The experimental flag '--expt-relaxed-constexpr' can be used to allow this.

/tmp/openmmTempKernel0x55cc511f2f30_7018.cu(1003): error: calling a constexpr __host__ function("fmin") from a __global__ function("computeBondedForces") is not allowed. The experimental flag '--expt-relaxed-constexpr' can be used to allow this.

2 errors detected in the compilation of "/tmp/tmpxft_00001b82_00000000-6_openmmTempKernel0x55cc511f2f30_7018.cpp1.ii".
Any ideas why this maybe? and why I am only having issues when using the CHARMM-GUI inputs?
Thank you for the help

[Peter Eastman]

Can you try the beta of OpenMM 7.5? THis error should be fixed in that version.

[Kyle Kihn]

Thank you for the help. I was able to download the beta cuda 7.5 and well anaconda to install the cuda toolkit in a new environment on one of my machines and it is working. However, on a machine that is very similar I did the same steps and it is still not working. This time there is a different error.
When I run:
python -m simtk.testInstallation
I get:
OpenMM Version: 7.5
Git Revision: cc3c4b54305fd535ddf26df15d9e1bb903e47

There are 4 Platforms available:
1 Reference - Sucessfully computed forces
2 CPU - Successfully computed forces
3 CUDA - Error computing forces with CUDA platform
4 OpenCL - Successfully computer forces

CUDA platform error: Error initializing FFT: 11
Thank you for the help!

[Peter Eastman]

That's a really strange error to get at that point. Error 11 is CUFFT_INVALID_DEVICE, which according to the documentation isn't even one of the result codes that function can return. Is it possible you have an inconsistency with the libraries you're linking to, for example that your CUDA toolkit isn't compatible with your driver?

[Kyle Kihn]

It looks like there were some inconstancies. I removed the cudatoolkit I had installed and openMM and reinstalled both on the machine. Both machines are working now.
Thank you for your help and time!