Open jchodera opened 4 years ago
Unless you're planning to run simulations with more than 2 million atoms on F@H, it's not important.
Is it OK if we get this bugfix in anyway? I'd like to get everything we know might impact AMD cards in here.
And I think there are some plans to do this with the SARS-CoV-2 spike protein...
It wouldn't be hard to backport this change, since it's a pretty small one and only affects one file. But before you hit this error, you'll first hit the one in https://github.com/openmm/openmm/pull/2806, and that's a much larger, much more complicated change. That one comes up at about 1.5 million atoms.
I think it will be much easier all around if we wait until we have a 7.5 based core before trying to run such large systems.
@peastman : OK, we can wait for 7.5.0.
Now that we have a benchmark project, we should be able to start testing cores based on 7.5.0 on a wide variety of projects once the OpenMM release is out.
Just making a note that we should backport https://github.com/openmm/openmm/pull/2829