2020 09 28 Open NEST Developer Video Conference

previous | list | next

Agenda

1. Welcome
2. Review of NEST User Mailing List
3. Project team round
4. In-depth discussion Benchmarking setups for CoSim and the NRP-NEST coupling (Kim Sontheimer and Cristian Jimenez-Romero)

Mailing list

• https://nest-simulator.org/mailinglist
• all threads answered since last meeting

Project team round

Here we discuss topics that need broader attention, for example questions that came up but are outside a single project's scope, larger planned changes/PRs that affect all teams or pending work that is blocked by external factors.

1. Models / NESTML
   • no report
2. PyNEST
   • please have a look at #1773, note your ideas/comments in the issue
3. Kernel
   • nothing to report
4. Installation
   • nothing to report, closed couple of issues, still looking for MacOS developers.
5. Infrastructure
   • no report
6. Documentation
   • no major updates, cleaup and PRs as usual
7. EBRAINS
   • nothing to report
8. (Feature) Random number generation
   • still waiting, currently no updates
9. (Feature) Automated Testing
   • no report

In-depth discussion: Benchmarking setups for CoSim and NRP–NEST coupling

Special guests are Kim Sontheimer and Cristian Jimenez-Romero of the SimLab Neuroscience (Jülich)

Benchmarking setups for CoSim

Introduction of the discussion by Kim Sontheimer

• In-transit coupling of NEST and Elephant, stream simulation output into a synchronous analysis with Elephant
• Current Implementation, MPI send spikeevents to in-transit server (sorting, transposition, …) coupled to down-stream analysis
• Plans are to scale up to test till which size of network the synchronous analysis can keep pace

Discussion

• neurons per rank are known
• clarification of "real-time" meaning, that simulation time advances at the same speed as the wall-clock.
• which scalable model could be used for testing?
  • hpc_benchmark long known
  • structured models available, but difficult to scale. need to keep reasonable activity
• real-time not relevant for analysis, but only for real-world applications like robotics, sound-click, …

NEST–NRP interface architecture

Introduction of the discussion by Cristian Jimenez-Romero

Discussion

• is a NEST simulator-simulator available for testing? – not necessary, computationally inexpensive simulations can create sufficiently large streams of data. NEST's dry-run mode is not yet usable for this in the new versions.
• NRP is free of real-time constraint. closed-loop simulations should be able to run as fast as possible in a simulated environment.
• interactions with NEST are coordinated also with NEST Desktop