Skip to content

Release 8.1.0
Compare
Choose a tag to compare
@github-actions github-actions released this 25 Mar 16:52
· 10 commits to e7633ff0b480a2b9ea61c4c3fb385cff4c48f7e0 since this release
8.1.0
047dd82

8.1.0

Release Date : 25-03-2022

What's New

  • Python wheels

    • Support for ARM (Aarch64) Platform wheels via CircleCI - Python 3.6 to 3.10 (#1676)
    • Apple M1 Platform wheels - Python 3.8 to 3.10 (#1649)
    • NVIDIA GPU enabled wheels
      • First iteration released under a separate package: neuron-gpu
        • Must use NVHPC 21.1 SDK when you compile your MOD files (#1657)
    • Improved CoreNEURON support (#1705)
      • Use vendor compilers like Intel, Cray or NVHPC for better performance when you compile MOD files using nrnivmodl.
    • Support for HPE MPT MPI library (#1498)
    • Linux wheels are now manylinux2014 (#1365)

  • MacOS universal2 package installer now available - Python 3.8 to 3.10

  • CoreNEURON

    • Tight integration for both in-memory and as well as file transfer mode
    • Extend CoreNEURON POINTER transfer to any RANGE variable in a NRN_THREAD (#1622)
    • Support for BEFORE/AFTER constructs in MOD file (#1581)

  • RXD

    • SaveState support added (#1586)
    • Dynamic extracellular tortuosity and volume fraction (#1260)
    • 3D support for importing multiple morphologies and for moving imported morphologies (#1147)

  • Dedicated CI repository nrn-build-ci to check builds on different distributions. Check this if you have build issues.

  • Documentation

    • Integrated various documentation & tutorials from neuron.yale.edu website to the readthedocs (#1674)
    • Added training videos, guides and a list of NEURON publications
    • CoreNEURON: usage, MOD files compatibility/migration, profiling and (GPU) performance benchmarking
    • RxD programmer's reference expanded and rewritten (#1680)
    • Propose changes directly from ReadTheDocs - click on Edit on GitHub

Breaking Changes

  • Models using _NrnThread in a VERBATIM block will fail now. Change it to NrnThread.See #1609.
  • Minimum CMake version supported is now 3.15.0 (#1408)
  • Python2 and Python3.5 support dropped
  • Legacy internal readline source code is removed, must have Readline installed (#1371)
  • Python wheels are now manylinux2014 (#1365)

Deprecations

  • NEURON is being migrated to a C++ codebase and beginning with the next release MOD files will be converted to C++ instead of C.
    This might break compatibility with some legacy MOD files containing VERBATIM blocks.
    We will update the NEURON documentation describing how one can make MOD files compatible with this change.
  • Python 3.6 Is End-of-Life - drop support after 8.1 release (#1651)

Bug Fixes

  • Can use ParallelContext.mpiabort_on_error(0) to say do not call to MPI_Abort on error (#1567)
  • OpenMPI initialisation crash: argv[argc] should be null. (#1682)
  • Avoid segfault if error during construction of POINT_PROCESS (#1627)
  • Fixes a 1D/3D voxelization problem where frusta are outside the 3D grid. (#1227)
  • Allow for two point (single section) SWC somas (#1144)
  • Fix for current response in 3D reaction-diffusion simulations (#1721)
  • rxdmath.abs no longer raises an exception (#1545)
  • Fixes for 3D reaction-diffusion grid alignment edge cases (#1471, #1227)
  • Update coreneuron : gpu bugfix for the BBP models (#1374)
  • Fix UndefinedBehaviourSanitizer errors. (#1518)
  • import3d_gui bugfix for HOC classes and top-level (#1159)

Improvements / Other Changes

  • Documentation
    • sphinx jupyter notebooks integration via nbsphinx (#1684)
    • Update CoreNEURON documentation. (#1650)
    • Building python wheels & managing NEURON Docker Images (#1469)
    • Short tutorial on using the profiler interface (#1453)
    • docs: Updated several rxd tutorials (#1061)
    • docs: WSL & Linux Python Wheel (#1496)
    • Update GPU profiling documentation (#1538)
  • Testing and coverage
    • Integration of tqperf model for CI and CoreNEURON testing (#1556)
    • Add olfactory bulb mode as external test (#1541)
    • Add reduced dentate model tests (#1515)
    • Add BBP channel-benchmark to CTest suite. (#1439)
    • Make neuronsimulator/nrntest part of the CI (#1429)
  • Refactorings
    • Move codebase to CPP (#763)
    • Remove all code related to NRN_REALTIME and NRN_DAQ (#1401)
    • Drop legacy hash implementations: nrnhash (#1403)
    • Remove absolute paths from generated code for ccache builds (#1574)
    • Using std::unordered_map as Gid2PreSyn/Int2TarList (#1568)
    • nrnivmodl succeeds when there are no mod files (#1583)
    • Fix UndefinedBehaviourSanitizer errors. (#1518)
    • Remove copy of readline (#1396)
    • Remove multisend specific for BlueGene (DMA) (#1294)
  • Improvements
    • Possible to prefer python from $PATH when building from source with CMake #1659
    • User can use MPI_LIB_NRN_PATH to provide MPI library on linux platform (#1172)
    • MPI Multisend available in NEURON by default (#1537)
    • modlunit: Add CHARGE keyword and map it to VALENCE. (#1508)
    • Allow user to specify dialog popup location. (#1487)
    • Allow python section.disconnect(). (#1451)
    • Implement BBSaveState for Python cells (#1355)
    • macos build_wheels: add /usr/x11 to CMAKE_PREFIX_PATH (#1565)
    • Add NRN_NMODL_CXX_FLAGS to facilitate cross compilation (#1174)
  • RXD
    • Faster convergence rate for surface voxel partial volume estimation (#1555)
    • Internal API for saving/restoring 3D voxelization (#1476)
    • Prevent RxD keeping objects alive (#1270, #1103, #1072)
    • Improved assignment of 3D voxels to segments (#1149)

Upgrade Steps

  • Linux wheels are now manylinux2014: upgrade your pip
  • Legacy internal readline source code is removed: install readline on your system
  • Python2 and Python3.5 support dropped: migrate to Python 3.7+ (3.6 is deprecated)

For the complete list of features and bug fixes, see Commits going into 8.1.0 in GitHub Issue #1719

PyPI

https://pypi.org/project/NEURON/8.1.0/#files

Assets 5
Loading