Issues Installing openMM 7.7

The functionality of OpenMM will (eventually) include everything that one would need to run modern molecular simulation.
POST REPLY
User avatar
Kyle Kihn
Posts: 29
Joined: Mon Apr 01, 2019 8:18 am

Issues Installing openMM 7.7

Post by Kyle Kihn » Tue May 17, 2022 12:58 pm

Hi All,
I am attempting to install openmm on a gpu accelerated linux machine. I installed openMM and cudatoolkit using the:
conda install -c conda-forge openmm cudatoolkit

which installed the following versions:
openmm 7.7.0 py39h15fbce5_1 conda-forge
cudatoolkit 11.6.0 habf752d_10 conda-forge

my nvcc --version gives:
Cuda compilation tools, release 10.0, V10.0.130

and nvidia-smi gives:
NVIDIA-SMI 470.103.01 Driver Version: 470.103.01 CUDA Version: 11.4

However when I run python -m openmm.testInstallation I get:
OpenMM Version: 7.7
Git Revision: 130124a3f9277b054ec40927360a6ad20c8f5fa6

There are 4 Platforms available:

1 Reference - Successfully computed forces
2 CPU - Successfully computed forces
3 CUDA - Error computing forces with CUDA platform
4 OpenCL - Successfully computed forces

CUDA platform error: Error launching CUDA compiler: 256
nvcc fatal : Value 'sm_86' is not defined for option 'gpu-architecture'


Median difference in forces between platforms:

Reference vs. CPU: 6.30846e-06
Reference vs. OpenCL: 6.76294e-06
CPU vs. OpenCL: 7.71916e-07

All differences are within tolerance.

Any ideas on how to resolve this issue? I have tried installing the cudatoolkit 10.0 but get the same errors.

Thanks for the help

User avatar
Peter Eastman
Posts: 2593
Joined: Thu Aug 09, 2007 1:25 pm

Re: Issues Installing openMM 7.7

Post by Peter Eastman » Tue May 17, 2022 1:06 pm

I think this indicates an incompatibility between the CUDA driver (11.4) and toolkit (11.6). You can upgrade your driver, or alternatively tell conda to install an older version:

Code: Select all

conda install -c conda-forge cudatoolkit=11.4

User avatar
Peter Eastman
Posts: 2593
Joined: Thu Aug 09, 2007 1:25 pm

Re: Issues Installing openMM 7.7

Post by Peter Eastman » Tue May 17, 2022 1:08 pm

I also wonder why it's using nvcc? Unless you specifically told it to do that with OPENMM_CUDA_COMPILER, it ought to be using the nvrtc compiler.

User avatar
Kyle Kihn
Posts: 29
Joined: Mon Apr 01, 2019 8:18 am

Re: Issues Installing openMM 7.7

Post by Kyle Kihn » Tue May 17, 2022 2:31 pm

Hi Peter,
Thank you so much for the help! I updated the CUDA Toolkit from Nvidia to 11.4, then changed my bashrc file to reflect this, and your suggested change.

After that I was able to install using simply conda install -c conda-forge openmm cudatoolkit=11.4

It seems to be working great now,
thanks for the help!

User avatar
Elsaid Mohamed
Posts: 26
Joined: Tue Jan 04, 2022 2:35 am

Re: Issues Installing openMM 7.7

Post by Elsaid Mohamed » Wed Jun 29, 2022 9:51 am

I was trying to reinstall open mm in my Linux pc, I got these errors after three hours:

~$ conda install -c conda-forge openmm=7.7.0
Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: \
Found conflicts! Looking for incompatible packages.
This can take several minutes. Press CTRL-C to abort.
Examining conflict for et_xmlfile snowballstemmer xlsxwriter wrapt datashader

Also, it does not work properly when I diid install it inside the new conda environment.

POST REPLY