Cache some charge increment calculations

[mattwthompson]

Description

For some systems, most of the time writing GROMACS files was spent adding up a bunch of charges. This is a shortcut to speed that up. Some rough checks using some of the slower toolkit examples:

$ python -m pytest --nbval-lax --dist loadscope -n logical --durations=20 ../openff-toolkit/examples/using_smirnoff_in_amber_or_gromacs ../openff-toolkit/examples/using_smirnoff_with_amber_protein_forcefield -p no:randomly > head.log

With e6f9f88:

============================= test session starts ==============================
platform darwin -- Python 3.11.11, pytest-8.3.4, pluggy-1.5.0
rootdir: /Users/mattthompson/software/openff-toolkit
configfile: pyproject.toml
plugins: cov-6.0.0, nbval-0.11.0, anyio-4.6.2.post1, xdist-3.6.1
created: 8/8 workers
8 workers [25 items]

.........................                                                [100%]
============================= slowest 20 durations =============================
119.34s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 9
82.66s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 3
49.39s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 8
49.14s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 10
48.09s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 5
29.25s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 6
8.10s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 0
7.50s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 4
6.49s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 1
3.43s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 3
2.03s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 2
0.85s setup    examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 0
0.85s setup    examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 0
0.69s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 6
0.53s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 1
0.45s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 2
0.37s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 8
0.30s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 12
0.27s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 11
0.16s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 9
======================== 25 passed in 396.25s (0:06:36) ========================

Compared to 0.4.0:

============================= test session starts ==============================
platform darwin -- Python 3.11.11, pytest-8.3.4, pluggy-1.5.0
rootdir: /Users/mattthompson/software/openff-toolkit
configfile: pyproject.toml
plugins: cov-6.0.0, nbval-0.11.0, anyio-4.6.2.post1, xdist-3.6.1
created: 8/8 workers
8 workers [25 items]

.........................                                                [100%]
============================= slowest 20 durations =============================
516.67s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 9
257.41s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 10
245.22s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 8
67.91s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 3
36.93s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 5
20.93s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 6
6.45s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 3
6.31s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 4
1.28s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 0
1.26s setup    examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 0
1.26s setup    examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 0
0.91s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 1
0.88s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 6
0.79s call     examples/using_smirnoff_with_amber_protein_forcefield/BRD4_inhibitor_benchmark.ipynb::Cell 2
0.59s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 1
0.52s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 12
0.27s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 2
0.27s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 11
0.22s call     examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 8
0.12s teardown examples/using_smirnoff_in_amber_or_gromacs/export_with_interchange.ipynb::Cell 13
======================= 25 passed in 1155.80s (0:19:15) ========================

Checklist

• Benchmark .to_gromacs on several systems
• Add tests
• Lint
• Update docstrings

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.64%. Comparing base (3e65e56) to head (e6f9f88).

Additional details and impacted files

[mattwthompson]

Some nice improvements for a couple of systems (GB3 down to ~2 seconds from 12-18)

This script should be gobbled up into a solution for #1020

import warnings

warnings.filterwarnings('ignore')

from openff.toolkit import ForceField, Molecule, Topology, Quantity
from IPython import get_ipython

gb3 = Topology.from_pdb("../proteinbenchmark/proteinbenchmark/data/pdbs/gb3-1P7E.pdb")
gb3.box_vectors = Quantity([10, 10, 10], "nanometer")

protein = ForceField('Protein-Specific-0.0.3.offxml').create_interchange(gb3)

caffeine = Molecule.from_smiles("CN1C=NC2=C1C(=O)N(C(=O)N2C)C")
caffeine.generate_conformers(n_conformers=1)
caffeine = caffeine.to_topology()
caffeine.box_vectors = Quantity([10, 10, 10], "nanometer")

ligand = ForceField("openff-2.2.1.offxml").create_interchange(caffeine)


print("Exporting ligand system:")
get_ipython().run_line_magic(
    "timeit",
    "ligand.to_gromacs('tmp')",
)

print("Exporting protein system:")
get_ipython().run_line_magic(
    "timeit",
    "protein.to_gromacs('tmp')",
)

$ ipython gmx_perf.py && git checkout upstream/main && ipython gmx_perf.py && git checkout v0.4.0 && ipython gmx_perf.py
Exporting ligand system:
9.48 ms ± 496 μs per loop (mean ± std. dev. of 7 runs, 100 loops each)
Exporting protein system:
2.15 s ± 72.7 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
HEAD is now at 3e65e567 Merge pull request #1072 from openforcefield/dependabot/github_actions/mamba-org/setup-micromamba-2
Exporting ligand system:
16.1 ms ± 277 μs per loop (mean ± std. dev. of 7 runs, 100 loops each)
Exporting protein system:
11.5 s ± 269 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
Previous HEAD position was 3e65e567 Merge pull request #1072 from openforcefield/dependabot/github_actions/mamba-org/setup-micromamba-2
HEAD is now at 7f93571f Merge pull request #1090 from openforcefield/fix-missed-decorator

Support for writing out LAMMPS data files will be removed
in mbuild 1.0.
See GMSO (https://github.com/mosdef-hub/gmso/tree/main/gmso/formats/lammpsdata) for
continued support for LAMMPS.


Exporting ligand system:
17.8 ms ± 3.29 ms per loop (mean ± std. dev. of 7 runs, 100 loops each)
Exporting protein system:
11.4 s ± 808 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)