Skip to content

Releases: arbor-sim/arbor

Arbor Library v0.4

15 Oct 14:31
79855b6
Compare
Choose a tag to compare

Release notes

Library

  • Moved from C++14 to C++17
    • Removed our hand-rolled versions of any, optional and variant.
  • Added std::expected equivalent for error handling.

Features

  • Added mechanism catalogues with mechanisms used by Allen and BBP models.
  • Removed support for spherical segments at the root of cable morphologies, and
    replaced the sample-based representation with a segment-based representation:
    • Morphologies are defined in terms of two-point segments.
    • Gaps are allowed between segments anywhere in a morphology.
  • Exposed the current time inside mechanisms.
  • Added support for NeuroML2 morphology descriptions.
  • Added a "stitch" morphology builder for constructing morphologies with
    cable sections that can connect to any location on their parent cable.
  • Replaced recipe probe API with more flexible API that allows for sampling
    not only voltages at single locations, but currents, ion species properties,
    and mechanism state variables at single locations or across an entire cell.
  • Added support for querying probe metadata from the simulation object.
  • Added new 'place_pwlin' C++ API for cell geometry queries.
  • Added support for loading Allen SDK cell model morphologies from SWC.
  • Added support for composing policies for creating compartments over sub-regions.

Documentation

  • Restructured documentation to have cleaner separation between high level descriptions
    of concepts and the C++ and Python APIs.
  • Added high level documentation for morphology descriptions, labels and cable cell
    construction.

Optimizations

  • Implemented memory optimizations for GPU matrix solver.
  • Added support for ARM SVE intrinsics in the vectorized CPU back end.

Bug Fixes

  • Fixed various modcc code generation errors.

Arbor Library v0.3

01 Apr 08:55
db44a3c
Compare
Choose a tag to compare

Arbor library version 0.3, tagged as v0.3

Arbor is a library for implementing performance portable network simulations of multi-compartment neuron models.

An installation guide and library documentation are available online at Read the Docs.

Submit a ticket if you have any questions or want help.

Change Log

Changes since v0.2:

  • Python wrapper with pip installation.
  • Replace the morphology specification API for more flexible cell building.
  • Flat descriptions of ion channel distribution and synapse placement.
  • Multi-compartment back end support for sub-branch mechanism distributions.
  • Improved NMODL support:
    • nonlinear kinetic schemes
    • linear system solution in initial conditions
    • many small features and bug fixes
  • Generic ion species.
  • Many optimizations and bug fixes.

Arbor Library v0.2.1

26 Aug 14:45
Compare
Choose a tag to compare

Arbor library version 0.2.1, tagged as v0.2.1

Arbor is a library for implementing performance portable network simulations of multi-compartment neuron models.

An installation guide and library documentation are available online at Read the Docs.

Submit a ticket if you have any questions or want help.

Release Notes.

Minor Update.

Arbor Library v0.2

04 Mar 17:46
6f70bfe
Compare
Choose a tag to compare

Arbor library version 0.2, tagged as v0.2

Arbor is a library for implementing performance portable network simulations of multi-compartment neuron models.

An installation guide and library documentation are available online at Read the Docs.

Submit a ticket if you have any questions or want help.

Some key features include:

  • Optimized back ends for CUDA, KNL, AVX2, ARM NEON intrinsics.
  • Asynchronous spike exchange that overlaps compute and communication.
  • Efficient sampling of voltage and current on all back ends.
  • Efficient implementation of all features on GPU.
  • Reporting of memory and energy consumption (when available on platform).
  • An API for addition of new cell types, e.g. LIF and Poisson spike generators.

Change Log

Changes since v0.1:

  • A new Hines matrix solver back end for the GPU that parallelises over cell
    branches, not cells, to increase the amount of parallelism. See #631.
  • Support for describing and simulating electrical gap junctions. See #661 #686.
  • An additional library libarborenv is now installed with useful
    helper functionality for managing the environment (e.g. detecting the number of available CPU cores). See #679.
  • Detection and allocation of GPUs to MPI ranks on systems with more than one GPU per node in libarborenv. See #659 and #654.
  • The miniapp example was removed and replaced with a simple single cell model that shows how to use morphologies. See #703 and #710.
  • Support for ARM NEON intrinsics. See #698.
  • Basic Python support. Full Python support is slated for v0.3. See #668.

Contributors

Nora Abi Akar
John Biddiscombe
Benjamin Cumming
Felix Huber
Marko Kabic
Vasileios Karakasis
Wouter Klijn
Anne Küsters
Alexander Peyser
Stuart Yates

Citation

DOI

Nora Abi Akar, John Biddiscombe, Benjamin Cumming, Felix Huber, Marko Kabic, Vasileios Karakasis, Wouter Klijn, Anne Küsters, Alexander Peyser, Stuart Yates. (2019, March 4). arbor-sim/arbor: Arbor Library v0.2 (Version v0.2). Zenodo. http://doi.org/10.5281/zenodo.2583709

Version 0.1: First release

12 Oct 08:32
Compare
Choose a tag to compare

Arbor Library

Arbor library version 0.1, tagged as v0.1

Arbor is a library for implementing performance portable network simulations of multi-compartment neuron models.

An installation guide and library documentation are available online at Read the Docs.

Submit a ticket if you have any questions or want help.

Some key features include:

* Optimized back ends for CUDA, KNL and AVX2 intrinsics.
* Asynchronous spike exchange that overlaps compute and communication.
* Efficient sampling of voltage and current on all back ends.
* Efficient implementation of all features on GPU.
* Reporting of memory and energy consumption (when available on platform).
* An API for addition of new cell types, e.g. LIF and Poisson spike generators.
* Validation tests against numeric/analytic models and NEURON.

Contributors

Nora Abi Akar
John Biddiscombe
Benjamin Cumming
Marko Kabic
Vasileios Karakasis
Wouter Klijn
Anne Küsters
Ivan Martinez
Alexander Peyser
Stuart Yates

Citation

DOI

If you use this version of Arbor, please cite it as Nora Abi Akar, John Biddiscombe, Benjamin Cumming, Marko Kabic, Vasileios Karakasis, Wouter Klijn, Anne Küsters, Ivan Martinez, Alexander Peyser, Stuart Yates. (2018, October 12). arbor-sim/arbor: Version 0.1: First release (Version v0.1). Zenodo. http://doi.org/10.5281/zenodo.1459679. The full citation is available in different formats on Zenodo.