Folding Forum

There is support in the OpenMM code, once Apple fixed a bug in its OpenCL implementation that used to exist. I would have to look up which version patch for OS X that fix was released in, that would be the minimum version possible to use.

Some researchers may even be running simulations directly using OpenMM if they happen to have a Mac with a powerful enough GPU, but I don't have any information on that The issue comes taking that OpenMM code and creating a folding core based on it to use on OS X. Also needed is the packaging to install any necessary support for the core to work on OS X. Then it would need to be tested before being deployed.

All of that is not easy, and would take resources from a limited pool of developers. There may also be other issues. For some of the examples look at the various topics about problems getting GPU folding on Linux and Windows going for some posting here.

JimboPalmer wrote:The science code that has to be the problem, is OpenMM.

http://openmm.org/about.html

I had thought that FAH uses gromacs. Is open a wrapper I'm not privy to the internal discussion around why some clients can use some GPU's and all others are left out. I'll do some forum diving since I'm stuck at home watching long c++ compiles

unsatchmo wrote:I had thought that FAH uses gromacs.

Gromacs is used for CPU folding, OpenMM is used for GPU folding.

unsatchmo wrote:I got this working. Here's what I did:

Confirming unsatchmo 's way to get GPU recognized on a 2010 MacPro 5,1 under OSX High Sierra .
As I have a NVIDIA Geforce GTX 770 GPU, I took the following additional steps:

1. Install CUDA for OSX

2. Code: Select all

   sudo ln -s /usr/local/cuda/lib/libcuda.dylib /usr/local/lib/libcuda.dylib

3. go to fahcontrol -> configure -> add a GPU Slot (leave erverything in default)
4. restart fahclient

   Code: Select all

   FAHClient --send-command shutdown

Now waiting for a WU to confirm it's really calculating

EDIT: missed the post above where it's told, that you will not get assigned a GPU WU on OSX. Leave the steps above for the records...