fix MPI cross-rank synaptic wiring

[ilkilic]

This PR fixes missing PSPs in MPI runs. Previously, it only created real synaptic connections when the presynaptic neuron was instantiated on the same rank, so cross-rank presynaptic sources were silently left unconnected and produced no PSPs. The fix uses ParallelContext gid-based wiring across ranks and adds a global gid mapping to support multi-pop circuits.

[codecov]

Codecov Report

❌ Patch coverage is 92.64706% with 15 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
bluecellulab/circuit_simulation.py	78.33%	13 Missing ⚠️
bluecellulab/reports/writers/compartment.py	66.66%	1 Missing ⚠️
tests/test_circuit_simulation_mpi.py	99.01%	1 Missing ⚠️

Files with missing lines	Coverage Δ
bluecellulab/connection.py	98.36% <100.00%> (+0.11%)	⬆️
bluecellulab/reports/manager.py	95.23% <100.00%> (ø)
bluecellulab/simulation/simulation.py	56.32% <ø> (-0.35%)	⬇️
tests/test_connection.py	100.00% <100.00%> (ø)
bluecellulab/reports/writers/compartment.py	86.02% <66.66%> (-0.80%)	⬇️
tests/test_circuit_simulation_mpi.py	99.01% <99.01%> (ø)
bluecellulab/circuit_simulation.py	85.71% <78.33%> (-0.88%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[darshanmandge]

Thanks, @ilkilic, for locating and fixing the bug! Have you compared the results of the reports—spikes, soma, compartment—for single-core and multicore runs? The content comparison of the H5 files should confirm if the results are consistent.

Please also compare the above reports for the same simulation config with the neurodamus reports.

[AurelienJaquier]

thanks, looks good to me!

[ilkilic]

Thanks, @ilkilic, for locating and fixing the bug! Have you compared the results of the reports—spikes, soma, compartment—for single-core and multicore runs? The content comparison of the H5 files should confirm if the results are consistent.

Please also compare the above reports for the same simulation config with the neurodamus reports.

Yes, I compared spikes/soma/compartment reports for single-core vs multicore and the h5 contents are identical.

I also compared against neurodamus and results are consistent, PSPs trigger in the same cells, though not 100% identical.

[darshanmandge]

PSPs trigger in the same cells, though not 100% identical.

Let's fix this soon in the next PR. Thanks.