Orientation restraints broken with OpenMP - Redmine #2223
I’m hazy on the details of why the issue occurs, or the form of the proper fix, but for posterity, the implementation of orientation restraints was buggy when run with OpenMP (so group scheme is fine, as is single-thread-per-MPI-rank with Verlet scheme).
For now, David has hacked that we go single threaded for orientation restraints during the bonded kernels. This will let us get the new test case merged and some other fixes testable.
(from redmine: issue id 2223, created on 2017-08-02 by mark.j.abraham, closed on 2017-08-03)
- Changesets:
- Revision 0be497b7 by David van der Spoel on 2017-08-02T17:14:38Z:
Worked around missing OpenMP implementation in orires.
The orientation restraint code is not aware of OpenMP threads
and uses some global information. By only running it on the
master node results are now independent of number of threads
used.
Fixes #2223
Change-Id: Ie86f4bd4e645fa71a58114950f6a297b5788e022
- Revision 71a17da6 by David van der Spoel on 2017-08-15T14:55:35Z:
Worked around missing OpenMP implementation in orires.
The orientation restraint code is not aware of OpenMP threads
and uses some global information. By only running it on the
master node results are now independent of number of threads
used.
Fixes #2223
Change-Id: Ie86f4bd4e645fa71a58114950f6a297b5788e022