Neurodamus

===============
**25 Nov 2022**

*New Features*
* [BBPBGLIB-957] Added RSS printing in multiple places

*Improvements*
* Avoid writing --report-conf to the sim.conf if reports are disabled
* Load balancing improvements for multiple populations

*Bug fixes*
* Clear the model after calling savestate()

==============
**09 Nov 2022**

*New Features*
* [BBPBGLIB-938] Clean Random123 objects in synapses
* [BBPBGLIB-950] Call malloc_trim to return free pages back to the OS when clearing up the model
* Shrink NEURON ArrayPools and call Python garbage collect when clearing up the model

==============
**07 Nov 2022**

*New Features*
* [NRN-111] Add support for incoming Datum changes in 9.0.0

==============
**04 Nov 2022**

*Improvements*
* [BBPP134-14] Support node files with '.sonata' extension

*Bug fixes*
* [BBPBGLIB-945] Properly display the exception messages when the simulation crashes

==============
**21 Oct 2022**

*New Features*
* Enable model stats printing in CoreNEURON
* Read new sonata keys from libsonata

* [BBPBGLIB-885] neuromodulation_dtc and neuromodulation_strength in connection_overrides
* [BBPBGLIB-915] deprecate minis_single_vesicle key from conditions
* [BBPBGLIB-913] deprecate forward_skip key
* [BBPBGLIB-920] add keys in run for additional seedings: stimulus_seed, ionchannel_seed, minis_seed, synapse_seed
* [BBPBGLIB-921] add series_resistance key in seclamp
* [BBPBGLIB-919] parse modifications

*Improvements*
* [BBPBGLIB-934] LoadBalance refactoring for multiple circuits
* Enable reading sonata circuit config with empty edge

*Bug fixes*
* [BBPBGLIB-933] Fix coreneuron multi-cycle model building for multiple circuits
* Fix for conflict with SHM File Transfer and --keep-build

==============
**07 Oct 2022**

*Improvements*
* Add a GapJunction unit test to showcase how it could be tested
* Improved Cell Managers API with `get_cell` (python cell) and `get_cellref` (hoc cellref)
* Make Load balancer use the TargetManager Python class
* [HPCTM-1600] Set SHM File Transfer by default + Improve model memory consumption estimates