GitLab

It may contain a garbage value, and if it is large then
updateNeighborListSize does not force reorder atoms after 25 steps
in extremal cases.

* enable atom subset in PDBReporter

* adds optional atomSubset argument to PDBReporter
* adds functions in PDBReporter which create a new topology
  and positions with chosen subset of atoms
* PDBReporter write PDB files with the subset topology and positions.

* refactor PDBReporter

* make _createTopologySubset a method that is called only the first time it is needed
* more efficient creation of subset positions
* check that atomSubset is ordered

* refactoring PDBReporter and add PDBx

* move checks on atomSubset to _createSubsetTopology
* copy periodic box vectors from topology to subsetTopology
* add atomSubset to PDBx reporter

* add bond subset to createSubsetTopology

* formatting changes to pdbreporter

* loop over atoms cleaned up
* put file opening inside parameter loop in
  TestPdbReporter::testinvalidSubsets to try and fix failing tests on windows

* add thorough tests to TestPdbReporter

* spelling changes
* Add tests for atom positions, elements, names etc
* attempt at fixing PyPy and Windows failing test cases

* fix spellings

* close output file before raising exceptions

* closes output files before raising exception in pdbreporter
createSubsetTopology
* changes assertVecAlmostEqual from a method to function to avoid
  repeating

* Use 768 instead of 384 threads in generic kernels

* Use 1536 instead of 1024 threads in force kernels.

Fix specifying a device will fail to run when multiple OpenCL platforms are available. Specifying `--verbose` prints the following exception:
```
Specified DeviceIndex but not OpenCLPlatformIndex. When multiple platforms are available, a platform index is needed to specify a device.
```

This PR should fix the docstring rendering of `Quantity`, which currently looks the following way: http://docs.openmm.org/latest/api-python/generated/openmm.unit.quantity.Quantity.html#openmm.unit.quantity.Quantity

* Fix misleading documentation of `Topology.bonds`

`Topology.bonds()` states that iterates over a tuple of `Atom` objects. While this is technically correct, as the `Bond` class is a named tuple, this also hides the fact, that `Bond` objects also provide the `order` and `type` attributes. Therefore, I propose to change the sentence, so that it uses the capitalized *Bonds* indicating that they are actually objects.

* Update docstring

- Appear to be copy and pasted from getPositions and were not removed

* Flushing optimization

* Remove unnecessary checks

* Profiling of OpenCL kernels

* Minor improvements to profiling

* update to use newer openmm-torch version

* openmm8 rc1 envs

* forgot to bump nnpops

* Update 05_cpu_platform.rst

* Update 04_platform_specifics.rst

* Use VkFFT for OpenCL

* Updated comments for OpenCLFFT3D

* Overhaul benchmarking script to add lots of new features

* Added another example

* Attempt at fixing failing pypy build

* Try using mambaforge

* Debugging

* A different approach to using mambaforge

* Try a separate environment file for pypy

* copy opc ions and opc*standard.xml from https://github.com/openmm/openmmforcefields/commit/b12c2e871b60275a10b3f8f56dfc2f6d5447591f

* auto-merged files

* move charges to residue definition

* remove unmerged files

* add docs

* rename U to U4+

* rename Ag+, Tl+, Gd3+ and I- residue names

* fix O-H bond length typo

* add opc3 constraints test

This initialization is needed to prevent segfault during object destruction in certain circumstances (e.g., when a `Force` is changed before calling `Context.reinitialize()` and this change causes the corresponding `ForceImpl` to throw an exception).

Fix #3790: Collision rate for MTS BAOAB Langevin integrator now correctly accounts for number of substeps (#3791)

* Fix #3790: Collision rate for MTS BAOAB Langevin integrator now correctly accounts for number of substeps. Fix suggested by Charlie Matthews (@c-matthews)

* Fix computation of total number of substeps

* Add test for MTS friction

* Fix typo

* Fix yet another typo

* Fix typo and check against analytical result

* Fix typo

* Fix typos

* Fix more typos

* Integrate MTSLangevinIntegrator for longer to allow thermalization

* Revert number of integrator steps

* Update TestIntegrators.py

Fixed a failing test case
Co-authored-by: Peter Eastman <peter.eastman@gmail.com>

Parallelize fitting 3D splines