BioNeuronCore

BioNeuronCore is the high-performance computational engine of the MindCopy platform for biologically realistic neural network simulation.

Technical Specifications

• Language: C++17 + SYCL 2020 (Intel oneAPI DPC++)
• CPU: GCC, static library libbiocore_gcc.a
• GPU: Intel DPC++, dynamic library libbiocore_gpu_backend.so
• Data model: POD, #pragma pack(1), USM-compatible

Operating Modes

• Interactive: via SimulationBridge (IPC Unix socket/TCP) — for Studio
• Batch: --batch --load-bion --ticks --params-json --seed --bexp-output — for Orchestrator
• As library: static link (CPU) or dlopen (GPU) — for direct embedding

Biophysics

• Hybrid HH+GHK: 59 built-in HH plugins, all currents via Goldman-Hodgkin-Katz equation
• 121 modifier types (ion channels, receptors, pumps, exchangers, sensors)
• Diffusion: Fick's law, 5 ions + 14 neurotransmitters + ATP + temperature
• Synaptic transmission: chemical (GPCR cascades, probabilistic release) + electrical (gap junctions)
• Plasticity: STDP (CaMKII/Calcineurin), BCM, PP1/DARPP-32, consolidation (6 events), gene expression (36 genes)
• Homeostasis: PMCA, NCX, KCC2, Na/K-ATPase, Ca²⁺ buffer
• Checkpoints: auto-save every N ticks, --resume-from, graceful shutdown (SIGTERM)

BION v6 Format

Binary neural network format: NetworkHeader (84B), Segment (988B), Modifier (44B), Synapse (824B), Nucleus (332B). Backward compatible v3→v4→v5→v6 with on-the-fly migration.

32 automated tests. Free for research use.