Add support for MDA minimize_vectors

[PardhavMaradani]

This PR adds support for MDAnalysis minimize_vectors as described in MDA Issue #5234. MDA PR #5235 adds the backend support for this function in MDAnalysis. Thanks

[PardhavMaradani]

Hi, I am unable to figure out the build error below. Could you please help me and let me know if there is something I am missing? I was able to build this locally on my ubuntu environment, which is how I was able to run the performance tests with a working version in the referenced MDA PR #5235. Thanks

[PardhavMaradani]

Hi, I am unable to figure out the build error below.

I noticed that the 'Read the Docs' build has been failing for quite some time. The rest of the tests didn't seem to run and I see the following notice for that workflow:

Could someone with access please re-enable this for tests to run again. Thanks

[PardhavMaradani]

Here are the benchmark results comparing against mda version (serial):

Show a full benchmark results run

(distopia) cmake-build/libdistopia$ ./bench --benchmark_filter="MinimizeVectorsBench"
2026-02-23T08:08:56+05:30
Running ./bench
Run on (8 X 4500 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x4)
  L1 Instruction 32 KiB (x4)
  L2 Unified 256 KiB (x4)
  L3 Unified 8192 KiB (x1)
Load Average: 2.13, 0.95, 0.65
***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
--------------------------------------------------------------------------------------------------------------------------------------
Benchmark                                                                            Time             CPU   Iterations UserCounters...
--------------------------------------------------------------------------------------------------------------------------------------
MinimizeVectorsBench<float>/MinimizeVectorsMDAOrthoFloat/10/0/0                   94.7 ns         94.7 ns      6775195 Per Vector=9.47334ns items_per_second=105.559M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDAOrthoFloat/100/0/0                   963 ns          963 ns       717287 Per Vector=9.62792ns items_per_second=103.865M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDAOrthoFloat/1000/0/0                 9700 ns         9700 ns        73683 Per Vector=9.69997ns items_per_second=103.093M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDAOrthoFloat/10000/0/0               95353 ns        95353 ns         6963 Per Vector=9.53532ns items_per_second=104.873M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDAOrthoFloat/100000/0/0             992317 ns       992301 ns          713 Per Vector=9.92301ns items_per_second=100.776M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDAOrthoFloat/1000000/0/0          10042317 ns     10042286 ns           65 Per Vector=10.0423ns items_per_second=99.5789M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDAOrthoFloat/10000000/0/0        108277702 ns    108274283 ns            5 Per Vector=10.8274ns items_per_second=92.358M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDAOrthoDouble/10/0/0                 91.4 ns         91.4 ns      7354795 Per Vector=9.14066ns items_per_second=109.401M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDAOrthoDouble/100/0/0                 906 ns          906 ns       770059 Per Vector=9.06389ns items_per_second=110.328M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDAOrthoDouble/1000/0/0              13134 ns        13134 ns        53279 Per Vector=13.1338ns items_per_second=76.1394M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDAOrthoDouble/10000/0/0            168584 ns       168557 ns         4142 Per Vector=16.8557ns items_per_second=59.3272M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDAOrthoDouble/100000/0/0          1778711 ns      1778314 ns          383 Per Vector=17.7831ns items_per_second=56.233M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDAOrthoDouble/1000000/0/0        18292648 ns     18290987 ns           36 Per Vector=18.291ns items_per_second=54.6717M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDAOrthoDouble/10000000/0/0      148940372 ns    148919329 ns            5 Per Vector=14.8919ns items_per_second=67.1505M/s
MinimizeVectorsBench<float>/MinimizeVectorsOrthoFloat/10/0/0                      16.3 ns         16.2 ns     43208169 Per Vector=1.62491ns items_per_second=615.418M/s
MinimizeVectorsBench<float>/MinimizeVectorsOrthoFloat/100/0/0                     86.4 ns         86.4 ns      8070018 Per Vector=864.065ps items_per_second=1.15732G/s
MinimizeVectorsBench<float>/MinimizeVectorsOrthoFloat/1000/0/0                     717 ns          717 ns       974746 Per Vector=717.257ps items_per_second=1.3942G/s
MinimizeVectorsBench<float>/MinimizeVectorsOrthoFloat/10000/0/0                   7669 ns         7668 ns        90964 Per Vector=766.842ps items_per_second=1.30405G/s
MinimizeVectorsBench<float>/MinimizeVectorsOrthoFloat/100000/0/0                 79084 ns        79082 ns         8730 Per Vector=790.822ps items_per_second=1.26451G/s
MinimizeVectorsBench<float>/MinimizeVectorsOrthoFloat/1000000/0/0              1790784 ns      1790715 ns          404 Per Vector=1.79071ns items_per_second=558.436M/s
MinimizeVectorsBench<float>/MinimizeVectorsOrthoFloat/10000000/0/0            22845507 ns     22844969 ns           29 Per Vector=2.2845ns items_per_second=437.733M/s
MinimizeVectorsBench<double>/MinimizeVectorsOrthoDouble/10/0/0                    20.7 ns         20.7 ns     35724392 Per Vector=2.07161ns items_per_second=482.716M/s
MinimizeVectorsBench<double>/MinimizeVectorsOrthoDouble/100/0/0                    131 ns          131 ns      5108576 Per Vector=1.30735ns items_per_second=764.904M/s
MinimizeVectorsBench<double>/MinimizeVectorsOrthoDouble/1000/0/0                  1251 ns         1251 ns       531390 Per Vector=1.25084ns items_per_second=799.465M/s
MinimizeVectorsBench<double>/MinimizeVectorsOrthoDouble/10000/0/0                12605 ns        12577 ns        54701 Per Vector=1.25775ns items_per_second=795.072M/s
MinimizeVectorsBench<double>/MinimizeVectorsOrthoDouble/100000/0/0              129167 ns       129166 ns         4503 Per Vector=1.29166ns items_per_second=774.2M/s
MinimizeVectorsBench<double>/MinimizeVectorsOrthoDouble/1000000/0/0            4042681 ns      4041971 ns          171 Per Vector=4.04197ns items_per_second=247.404M/s
MinimizeVectorsBench<double>/MinimizeVectorsOrthoDouble/10000000/0/0          46757814 ns     46754429 ns           14 Per Vector=4.67544ns items_per_second=213.883M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDATriclinicFloat/10/0/0                407 ns          407 ns      1725397 Per Vector=40.713ns items_per_second=24.5622M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDATriclinicFloat/100/0/0              4028 ns         4028 ns       172412 Per Vector=40.2762ns items_per_second=24.8286M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDATriclinicFloat/1000/0/0            49541 ns        49540 ns        13663 Per Vector=49.5405ns items_per_second=20.1855M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDATriclinicFloat/10000/0/0          579736 ns       579721 ns         1165 Per Vector=57.9721ns items_per_second=17.2497M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDATriclinicFloat/100000/0/0        5813517 ns      5813530 ns          118 Per Vector=58.1353ns items_per_second=17.2013M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDATriclinicFloat/1000000/0/0      59057284 ns     59056222 ns           11 Per Vector=59.0562ns items_per_second=16.933M/s
MinimizeVectorsBench<float>/MinimizeVectorsMDATriclinicFloat/10000000/0/0    629116040 ns    629096172 ns            1 Per Vector=62.9096ns items_per_second=15.8958M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDATriclinicDouble/10/0/0              388 ns          388 ns      1800398 Per Vector=38.7783ns items_per_second=25.7876M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDATriclinicDouble/100/0/0            3839 ns         3839 ns       182470 Per Vector=38.3903ns items_per_second=26.0482M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDATriclinicDouble/1000/0/0          51414 ns        51409 ns        13338 Per Vector=51.409ns items_per_second=19.4519M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDATriclinicDouble/10000/0/0        626206 ns       626194 ns         1096 Per Vector=62.6194ns items_per_second=15.9695M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDATriclinicDouble/100000/0/0      6303879 ns      6303962 ns          109 Per Vector=63.0396ns items_per_second=15.863M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDATriclinicDouble/1000000/0/0    64112752 ns     64112265 ns           10 Per Vector=64.1123ns items_per_second=15.5976M/s
MinimizeVectorsBench<double>/MinimizeVectorsMDATriclinicDouble/10000000/0/0  717845553 ns    717796220 ns            1 Per Vector=71.7796ns items_per_second=13.9315M/s
MinimizeVectorsBench<float>/MinimizeVectorsTriclinicFloat/10/0/0                   129 ns          129 ns      5393738 Per Vector=12.9301ns items_per_second=77.3392M/s
MinimizeVectorsBench<float>/MinimizeVectorsTriclinicFloat/100/0/0                  796 ns          796 ns       871357 Per Vector=7.96363ns items_per_second=125.571M/s
MinimizeVectorsBench<float>/MinimizeVectorsTriclinicFloat/1000/0/0                7579 ns         7578 ns        92136 Per Vector=7.57812ns items_per_second=131.959M/s
MinimizeVectorsBench<float>/MinimizeVectorsTriclinicFloat/10000/0/0              75614 ns        75612 ns         9297 Per Vector=7.56123ns items_per_second=132.254M/s
MinimizeVectorsBench<float>/MinimizeVectorsTriclinicFloat/100000/0/0            816490 ns       816429 ns          918 Per Vector=8.16429ns items_per_second=122.485M/s
MinimizeVectorsBench<float>/MinimizeVectorsTriclinicFloat/1000000/0/0          7795943 ns      7795841 ns           87 Per Vector=7.79584ns items_per_second=128.274M/s
MinimizeVectorsBench<float>/MinimizeVectorsTriclinicFloat/10000000/0/0        82889566 ns     82888874 ns            8 Per Vector=8.28889ns items_per_second=120.643M/s
MinimizeVectorsBench<double>/MinimizeVectorsTriclinicDouble/10/0/0                 188 ns          188 ns      3713353 Per Vector=18.8489ns items_per_second=53.0534M/s
MinimizeVectorsBench<double>/MinimizeVectorsTriclinicDouble/100/0/0               1530 ns         1530 ns       456905 Per Vector=15.2972ns items_per_second=65.3715M/s
MinimizeVectorsBench<double>/MinimizeVectorsTriclinicDouble/1000/0/0             14748 ns        14748 ns        45064 Per Vector=14.7481ns items_per_second=67.8053M/s
MinimizeVectorsBench<double>/MinimizeVectorsTriclinicDouble/10000/0/0           149509 ns       149511 ns         4580 Per Vector=14.9511ns items_per_second=66.8847M/s
MinimizeVectorsBench<double>/MinimizeVectorsTriclinicDouble/100000/0/0         1502468 ns      1502380 ns          470 Per Vector=15.0238ns items_per_second=66.5611M/s
MinimizeVectorsBench<double>/MinimizeVectorsTriclinicDouble/1000000/0/0       15557782 ns     15557979 ns           44 Per Vector=15.558ns items_per_second=64.2757M/s
MinimizeVectorsBench<double>/MinimizeVectorsTriclinicDouble/10000000/0/0     183798640 ns    183800616 ns            3 Per Vector=18.3801ns items_per_second=54.4068M/s
(distopia) cmake-build/libdistopia$ 

Here are some comparative results using googlebench/tools/compare.py tool for a different run:

Comparing MinimizeVectorsBench<.*>/MinimizeVectorsMDA to MinimizeVectorsBench<.*>/MinimizeVectors (from results.json)

Benchmark                          Time             CPU      Time Old      Time New       CPU Old       CPU New
---------------------------------------------------------------------------------------------------------------
OrthoFloat/10/0/0               -0.7766         -0.7764            80            18            80            18
OrthoFloat/100/0/0              -0.8897         -0.8896           798            88           797            88
OrthoFloat/1000/0/0             -0.9081         -0.9064          7956           731          7943           743
OrthoFloat/10000/0/0            -0.9051         -0.9021         80056          7594         79937          7829
OrthoFloat/100000/0/0           -0.9009         -0.8980        805557         79805        804438         82024
OrthoFloat/1000000/0/0          -0.7773         -0.7716       8334606       1855913       8323315       1901048
OrthoFloat/10000000/0/0         -0.7347         -0.7294      89289238      23689467      89182779      24130552
OrthoDouble/10/0/0              -0.7688         -0.7648            83            19            83            20
OrthoDouble/100/0/0             -0.8338         -0.8312           783           130           782           132
OrthoDouble/1000/0/0            -0.8968         -0.8954         11482          1184         11469          1200
OrthoDouble/10000/0/0           -0.9080         -0.9068        142127         13073        141985         13227
OrthoDouble/100000/0/0          -0.9071         -0.9060       1428396        132695       1427016        134101
OrthoDouble/1000000/0/0         -0.7107         -0.7079      14723204       4259499      14710036       4296081
OrthoDouble/10000000/0/0        -0.7169         -0.7150     174023083      49264850     173889632      49553795
TriclinicFloat/10/0/0           -0.6884         -0.6896           442           138           444           138
TriclinicFloat/100/0/0          -0.8071         -0.8077          4395           848          4414           849
TriclinicFloat/1000/0/0         -0.8498         -0.8502         53472          8033         53683          8042
TriclinicFloat/10000/0/0        -0.8718         -0.8721        627857         80511        630109         80595
TriclinicFloat/100000/0/0       -0.8722         -0.8725       6294336        804605       6314923        805399
TriclinicFloat/1000000/0/0      -0.8703         -0.8706      63830021       8277742      64024160       8285230
TriclinicFloat/10000000/0/0     -0.8624         -0.8627     684613430      94179948     686476579      94256904
TriclinicDouble/10/0/0          -0.5339         -0.5347           428           199           429           200
TriclinicDouble/100/0/0         -0.6196         -0.6201          4265          1622          4274          1624
TriclinicDouble/1000/0/0        -0.7127         -0.7131         55492         15942         55600         15953
TriclinicDouble/10000/0/0       -0.7667         -0.7670        687784        160427        688997        160528
TriclinicDouble/100000/0/0      -0.7683         -0.7686       6934144       1606387       6945437       1607214
TriclinicDouble/1000000/0/0     -0.7539         -0.7541      70476422      17343621      70581484      17353843
TriclinicDouble/10000000/0/0    -0.7405         -0.7407     780051590     202449141     781122607     202570979
OVERALL_GEOMEAN                 -0.5747         -0.5732             0             0             0             0

MDAnalysis's own asv benchmark results comparing against different backends (including this version of distopia) can be found in this comment of MDAnalysis/mdanalysis#5235. Thanks

[orbeckst]

I re-enabled the workflow. Try pushing something to get the tests to run.

[orbeckst]

@hmacdope could you review, please?

[PardhavMaradani]

I re-enabled the workflow. Try pushing something to get the tests to run.

I pushed an empty commit. It seems like the workflow requires an approval to run. Thanks

[PardhavMaradani]

About the test failures:

• readthedocs has been failing for a while - looks like that build it not cloning the googlebench/test/higway submodules
• macos build failures don't seem related to changes in this PR - maybe macos builds are broken #186 is for these issues?

[PardhavMaradani]

The macos build failures show the following errors:

/Users/runner/work/distopia/distopia/libdistopia/googlebench/include/benchmark/benchmark.h:1446:30: error: '__COUNTER__' is a C2y extension [-Werror,-Wc2y-extensions]
 1446 | #if defined(__COUNTER__) && (__COUNTER__ + 1 == __COUNTER__ + 0)
      |                              ^
/Users/runner/work/distopia/distopia/libdistopia/googlebench/include/benchmark/benchmark.h:1446:49: error: '__COUNTER__' is a C2y extension [-Werror,-Wc2y-extensions]
 1446 | #if defined(__COUNTER__) && (__COUNTER__ + 1 == __COUNTER__ + 0)
      |                                                 ^
2 errors generated.

These seem to be fixed in the latest (v1.9.5) of google benchmark (through their PR google/benchmark#2108). Our submodule was pointing to an older commit ca8d0f7 (which was showing version 1.8.3 in the failed logs). I now updated the benchmark submodule to v1.9.5 (and verified locally). Could you please approve the workflow so we can verify that the builds and tests can run. Thanks

[orbeckst]

@PardhavMaradani I changed the repo settings for actions. Hopefully, you won't need approval for re-running CI. If it's still necessary please let me know.

[PardhavMaradani]

Updating google benchmark to v1.9.5 fixed the macos build issues. The same commit (5b0fe5a) showed a single test failure in pip-install (macos-latest, 3.14). They ran fine for other versions of python on macos and also for all python versions on ubuntu.

Capturing the relevant sections of the logs here so this is preserved:

Details

distopia/tests/test_distopia.py ........................................ [  5%]
........................................................................ [ 16%]
........................................................................ [ 26%]
........................................................................ [ 36%]
........................................................................ [ 47%]
Fatal Python error: Aborted

Current thread 0x0000000[20](https://github.com/MDAnalysis/distopia/actions/runs/22629140744/job/65617076210?pr=188#step:8:21)7122240 (most recent call first):
..............................................
  File "/Library/Frameworks/Python.framework/Versions/3.14/lib/python3.14/site-packages/_pytest/logging.py", line 388 in reset
  File "/Library/Frameworks/Python.framework/Versions/3.14/lib/python3.14/site-packages/_pytest/logging.py", line 8[25](https://github.com/MDAnalysis/distopia/actions/runs/22629140744/job/65617076210?pr=188#step:8:26) in _runtest_for
  File "/Library/Frameworks/Python.framework/Versions/3.14/lib/python3.14/contextlib.py", line 141 in __enter__
  File "/Library/Frameworks/Python.framework/Versions/3.14/lib/python3.14/site-packages/_pytest/logging.py", line 849 in pytest_runtest_call
.....
.....
Current thread's C stack trace (most recent call first):
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at _Py_DumpStack+0x48 [0x1040d0fd8]
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at faulthandler_dump_c_stack+0x58 [0x1040e5254]
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at faulthandler_fatal_error+0x17c [0x1040e5150]
  Binary file "/usr/lib/system/libsystem_platform.dylib", at _sigtramp+0x38 [0x1991c16a4]
  Binary file "/usr/lib/system/libsystem_pthread.dylib", at pthread_kill+0x128 [0x199187848]
  Binary file "/usr/lib/system/libsystem_c.dylib", at abort+0x7c [0x1990909e4]
  Binary file "/usr/lib/system/libsystem_malloc.dylib", at malloc_vreport+0x37c [0x198f94174]
  Binary file "/usr/lib/system/libsystem_malloc.dylib", at malloc_zone_error+0x64 [0x198fbf954]
  Binary file "/usr/lib/system/libsystem_malloc.dylib", at nanov2_guard_corruption_detected+0x2c [0x198fad494]
  Binary file "/usr/lib/system/libsystem_malloc.dylib", at nanov2_allocate_outlined+0x1cc [0x198fad460]
  Binary file "/usr/lib/system/libsystem_malloc.dylib", at nanov2_malloc_type+0x1d8 [0x198fabbac]
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at _PyObject_Malloc+0x6[58](https://github.com/MDAnalysis/distopia/actions/runs/22629140744/job/65617076210?pr=188#step:8:59) [0x103e99dec]
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at stringio_new+0x34 [0x10412ec1c]
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at type_call+0x[60](https://github.com/MDAnalysis/distopia/actions/runs/22629140744/job/65617076210?pr=188#step:8:61) [0x103eb6e7c]
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at _PyEval_EvalFrameDefault+0x2b62c [0x103fa83a4]
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at gen_iternext+0x88 [0x103e25be8]
  Binary file "/Library/Frameworks/Python.framework/Versions/3.14/Python", at builtin_next+0x48 [0x103f76e24]
.....
  <truncated rest of calls>

Extension modules: numpy._core._multiarray_umath, numpy.linalg._umath_linalg, distopia.pydistopia (total: 3)
/Users/runner/work/_temp/058c299c-045d-4770-ab95-3e9fa1b7dd5c.sh: line 1:  6331 Abort trap: 6           pytest distopia/tests
Error: Process completed with exit code 134.

Adding a -p no:logging to pytest as part of commit 0678ae3 above fixed the crash - this seems to indicate that this is a pytest issue. (Added -v also temporarily to check if the crash was due to any specific test of ours) I can undo this after your feedback on this. (not sure if I should debug this further) Thanks