Skip to content

SC-NEUROCORE — Module Integration Reference

Status: Historical integration inventory; use the live system map and focused module evidence for current status Snapshot date: 2026-04-16 Snapshot tests: 1,173 across 19 modules Languages: Python, Go, Rust, SystemVerilog

Overview

This document records the April 2026 migration from the research sandbox into the SC-NEUROCORE tree. Paths, tests, and readiness have continued to evolve; this inventory must not be used as a live certification or production-readiness claim.

At that snapshot, modules were:

  1. Copied (not moved) into src/sc_neurocore/ and tests/
  2. Given clean module names (no phase prefixes)
  3. Verified with full regression from the new locations
  4. Documented with class/function inventories for downstream documentation

Module Registry

Safety & Certification

safety_cert — fail-closed safety-evidence organisation

  • Source: ten responsibility modules plus the compatibility facade under src/sc_neurocore/safety_cert/
  • Tests: tests/test_safety_cert/ — 456 focused tests
  • Key Classes:
  • Requirement and TraceabilityMatrix — explicit implementation and verification links
  • FailureMode, FMEDA, and ReliabilityMetrics — arithmetic over caller-supplied FIT/DC records
  • FormalProperty and FormalProofCertificate — full-field evidence hashing and reports
  • CertificationGenerator and CertificationPackage — fail-closed assembly and atomic materialisation
  • EvidenceItem and EvidenceBag — path-safe SHA-256 manifests and on-disk verification

This library organises evidence. It does not grant certification, approval, or conformity.

fault_injection — Resilience Testing

  • Source: src/sc_neurocore/fault_injection/fault_injection.py
  • Tests: tests/test_fault_injection/test_fault_injection.py — 22 tests
  • Key Classes:
  • FaultInjector — Configurable bit-flip / stuck-at fault injector
  • FaultModel — Fault type definitions (SEU, MBU, stuck-at)
  • RadiationProfile — Space/terrestrial radiation models
  • ResilienceBenchmark — Automated resilience scoring

ASIC & Hardware

asic_flow — Multi-PDK ASIC Generation

  • Sources: src/sc_neurocore/asic_flow/{pdk,design,decks,signoff,constraints,estimation,flow,hierarchy,readiness}.py
  • Compatibility facade: src/sc_neurocore/asic_flow/asic_flow.py
  • Tests: tests/test_asic_flow/ — 100 focused tests
  • Key Classes:
  • PDKConfig, OpenSourcePDKResolver — Sky130/GF180MCU and custom PDK inputs
  • ASICFlowGenerator, ASICFlowBundle — deterministic decks and evidence manifests
  • SynthesisGenerator, FloorplanGenerator, PlaceRouteGenerator — Yosys/OpenROAD decks
  • SignoffGenerator, MultiCornerAnalysis, OCVConfig — signoff scripts and conditions
  • TapeOutChecklist — evidence-derived readiness state

uvm_gen — UVM Testbench Generator

  • Compatibility facade: src/sc_neurocore/uvm_gen/uvm_gen.py
  • Sources: src/sc_neurocore/uvm_gen/{_rtl,_config,_benchmark,_component_emitters,_harness_emitters,_generator}.py
  • Tests: tests/test_uvm_gen/ — 82 focused tests
  • Key Classes:
  • UVMGenerator — Full UVM testbench emission
  • RTLModule — RTL module abstraction
  • StimulusConfig — Constrained-random stimulus
  • CoverageSpec — Functional coverage specification
  • ScoreboardConfig — Self-checking scoreboard
  • Evidence boundary: Stable Python API and byte-exact deterministic text; emitted vendor-UVM/formal execution remains draft and is not promoted by the responsibility split.

chiplet — Multi-Die Chiplet Generator

  • Sources: src/sc_neurocore/chiplet/{topology,routing,thermal,rtl,link_protocols,power,partition}.py
  • Compatibility facade: src/sc_neurocore/chiplet/chiplet_gen.py
  • Tests: tests/test_chiplet/test_chiplet_*.py — 172 focused tests
  • Key Classes:
  • ChipletGenerator — Die-level generation with UCIe/BoW links
  • ChipletTopology — Die-to-die topology and interposer-link model
  • DieThermal / PackageThermalReport — Steady-state and transient thermal analysis
  • RoutingTable / PartitionAssignment — Routing and graph-to-die placement

nir_bridge — NIR-to-Silicon Interchange

  • Hardware-graph sources: src/sc_neurocore/nir_bridge/neuron_graph.py is the historical facade over seven acyclic responsibility modules for graph contracts, node classification, hierarchy inlining, dense lowering, connection traversal, metadata, and conversion orchestration
  • Focused tests: tests/test_nir_bridge/ owns parser, export, execution, framework interop, SC-NIR/FPGA integration, graph architecture, and malformed lowering contracts
  • Pipeline: from_nir() parses an external NIR graph; from_scnetwork() produces a hardware NeuronGraph; compile_network_to_fpga() consumes the graph without an adapter or duplicate intermediate representation
  • Compatibility: NeuronSpec, ConnectionSpec, HierarchyInstanceSpec, NeuronGraph, and from_scnetwork retain their historical qualified names and serialisation paths
  • Boundaries: nested graphs are flattened with hierarchy provenance; flatten/scale/delay/threshold metadata is preserved or rejected when it cannot be represented exactly; this orchestration surface has no numerical Rust, Go, Julia, or Mojo kernel counterpart

Physical Substrates

spintronic — Magnetic-Domain SC Mapper

  • Source: src/sc_neurocore/spintronic/spintronic_mapper.py
  • Tests: tests/test_spintronic/test_spintronic_mapper.py — 66 tests
  • Key Classes:
  • SpintronicMapper — Maps SC networks to MTJ arrays
  • MagneticDomainSim — Micromagnetic co-simulation bridge
  • SpinTorqueModel — STT/SOT write models
  • TMRCalculator — Tunnelling magnetoresistance estimation

memristor — Memristor Crossbar Mapper

  • Source: src/sc_neurocore/memristor/memristor_mapper.py
  • Tests: tests/test_memristor/test_memristor_mapper.py — 70 tests
  • Key Classes:
  • CrossbarArray — Memristive crossbar abstraction
  • ConductanceModel — Non-linear conductance model
  • AgingSimulator — Device degradation simulation
  • CrossbarEstimator — Power/area estimation

analog_bridge — Stochastic-to-Analog Bridge

  • Source: src/sc_neurocore/analog_bridge/analog_bridge.py
  • Tests: tests/test_analog_bridge/test_analog_bridge.py — 27 tests
  • Key Classes:
  • AnalogBridge — SC-to-analog conversion layer
  • DACModel — Digital-to-analog output model
  • ADCModel — Analog-to-digital input model

optics — Photonic compilation and crosstalk analysis

  • Compatibility facade: src/sc_neurocore/optics/photonic_emitter.py
  • Sources: seven bounded _photonic_* responsibility modules for types, conversion, FDTD, compilation, Meep, crosstalk, and bitstream emission
  • Focused verification: 102 tests, with exact 100% coverage over 702 statements and 214 branches in the facade and responsibility modules
  • Key Classes:
  • PhotonicCompiler and CompilationResult — validated optical compilation, netlist, FDTD, and GDSII orchestration
  • FDTDSolver and FDTD2DSolver — NumPy Yee-grid simulation
  • MeepAdapter — fail-closed optional pymeep bridge
  • CrosstalkModel and WaveguidePair — coupled-mode crosstalk contracts
  • Backend boundary: the engine plus standalone Rust, Go, Julia, and Mojo mirrors cover only crosstalk mathematics. Python exclusively owns FDTD, Meep, compilation, netlist, GDSII, and filesystem behaviour.

Exascale & Runtime

hypervisor — Neuromorphic Multi-Tenant Hypervisor

  • Source: src/sc_neurocore/hypervisor/hypervisor.py
  • Tests: tests/test_hypervisor/test_hypervisor.py — 78 tests
  • Key Classes:
  • Hypervisor — Multi-tenant SC workload manager
  • Tenant — Isolated workload with resource quotas
  • ResourceAllocator — FPGA tile / memory partitioning
  • PreemptionPolicy — Priority-based preemption
  • HealthMonitor — Watchdog and heartbeat monitoring

digital_twin/twinsync — Digital Twin Synchronization

  • Source: src/sc_neurocore/digital_twin/twinsync.py
  • Tests: tests/test_twinsync/test_twinsync.py — 72 tests
  • Key Classes:
  • TwinSession — Time-warp optimistic simulation
  • NullMessageOptimizer — Conservative sync lookahead
  • DeltaCheckpoint — Memory-efficient state diffs
  • ReplayVerifier — Deterministic replay verification
  • DriftAutoCorrector — Real-time drift compensation
  • CheckpointAuditChain — SHA-256 tamper-evident chain
  • TwinFederation — Multi-twin GVT coordination

Frontiers

evo_substrate — Self-Replicating Evolutionary Substrate

  • Source: src/sc_neurocore/evo_substrate/ — 156-line compatibility facade over 14 responsibility modules; no implementation file exceeds 304 lines
  • Tests: tests/test_evo_substrate/ — 160 tests across 18 focused modules, including architecture, Rust/Julia/Go/Mojo parity, and runtime-fault contracts
  • Boundaries: genome/organism own state; variation, fitness, speciation, and selection own search operators; safety, deployment, and emission own fail-closed hardware boundaries; replication composes the acyclic dependency graph
  • Key Classes:
  • Genome, TopologyGene, NeuronGene, PlasticityGene — Genetic encoding
  • MutationEngine — Point, structural, duplication, swap mutations
  • CrossoverEngine — Multi-point crossover
  • FitnessEvaluator — Multi-objective fitness
  • ReplicationEngine — Generation-level evolution loop
  • OrganismEmitter — NIR + Verilog organism emission
  • SafetyBounds — Mutation space constraints
  • TileDeploymentTracker — FPGA tile allocation
  • HallOfFame — Historical best tracking
  • IslandModel — Multi-deme with migration
  • NoveltyArchive — Behavioural novelty search
  • ExtinctionDetector — Stagnation-triggered mass extinction
  • CoevolutionArena — Predator-prey dynamics
  • FormalSafetyGuard — Pre-deployment validation
  • TournamentSelector — Configurable selection pressure
  • ParetoFront — NSGA-II style non-dominated front
  • AgeRegulator — Maximum lifespan enforcement
  • BloatPenalizer — Genome complexity regularization
  • CPPNGenome — Compositional Pattern Producing Networks
  • HWFitnessCollector — FPGA execution feedback
  • EvoStatisticsTracker — Per-generation analytics
  • ComplexityTracker — Open-ended complexity metric

meta_plasticity — Mutable Plasticity Rules

  • Source: src/sc_neurocore/meta_plasticity/meta_plasticity.py
  • Tests: tests/test_meta_plasticity/test_meta_plasticity.py — 72 tests
  • Key Classes:
  • MetaPlasticityEngine — Self-modifying learning rules
  • PlasticityRule — Parameterised STDP/triplet rules
  • RuleEvolver — Evolutionary plasticity rule search
  • ConsolidationScheduler — Memory consolidation timing

autonomous_learning — High-Performance Online Plasticity Engine

  • Source: crates/autonomous_learning/, src/sc_neurocore/_native/, src/sc_neurocore/accel/go/autonomous_learning/
  • Tests: tests/test_learning/ — 3 tests (Parity & Unit)
  • Key Classes:
  • RustPlasticityRule — FFI Python class mapping STDP/BCM/R-STDP to Rust backend
  • LearningBridgeAccel — Julia C-FFI wrapper
  • autonomous_learning.PlasticityRule (Go) — Go Cgo wrapper

bioware — Biological-Hardware Interface

  • Source: src/sc_neurocore/bioware/ — historical facade plus nine responsibility modules and shared validation
  • Tests: tests/test_bioware/ — 200 tests
  • Key Classes:
  • BioHybridSession — Transactional one-frame MEA→AER→SC→opto orchestration
  • SpikeDetector / SpikeSorter — MAD threshold detection and optional PCA/K-Means sorting
  • MEAToAERTranscoder / AERToSCConverter — Bounded AER epoch and deterministic LFSR encoding
  • SCToOptoEncoder — Area-aware optical-power budgeting
  • BiologicalSTDP / BCMPlasticity / HomeostaticPlasticity — Caller-managed plasticity adapters

federated — Federated SC Learning

  • Source: src/sc_neurocore/federated/federated_sc.py
  • Tests: tests/test_federated/test_federated_sc.py — 93 tests
  • Key Classes:
  • FederatedCoordinator — Privacy-preserving aggregation
  • SecureAggregator — Secure multi-party computation
  • DifferentialPrivacy — DP-SGD noise injection
  • ModelCompressor — Communication-efficient compression

bci_studio — BCI Closed-Loop Control

  • Source: src/sc_neurocore/bci_studio/bci_primitives.py, bci_studio.py
  • Tests: tests/test_bci/ — 32 tests
  • Key Classes:
  • BCIClosedLoopEngine — Real-time neural decoding
  • StimulusGenerator — Closed-loop stimulation
  • SafetyMonitor — Charge density limiter

Unification

explainability — SC Explainability Tools

  • Source: src/sc_neurocore/explainability/explainability.py
  • Tests: tests/test_explainability/test_explainability.py — 71 tests
  • Key Classes:
  • ExplainabilityEngine — Bitstream-level explanation generation
  • CausalAttributor — Causal attribution for SC decisions
  • FormalPropertyLink — Formal verification anchoring

neuro_symbolic — Predictive Coding Primitives

  • Source: src/sc_neurocore/neuro_symbolic/predictive_coding.py
  • Tests: tests/test_neuro_symbolic/test_predictive_coding.py — 34 tests
  • Key Classes:
  • PredictiveCodingLayer — Hierarchical prediction error
  • SymbolEncoder — Hyperdimensional symbol binding
  • VerifiableInference — Formally verifiable reasoning

stochastic_doctor — Bitstream Diagnostics

  • Source: src/sc_neurocore/stochastic_doctor/diagnostics.py
  • Tests: tests/test_stochastic_doctor/test_diagnostics.py — 16 tests
  • Key Classes:
  • BitstreamDoctor — Health diagnostics engine
  • CorrelationDetector — Inter-stream correlation analysis
  • BitstreamAuditReport — Full audit trail

model_zoo — Auto-Verilog Model Zoo

  • Source: src/sc_neurocore/model_zoo/model_zoo.py
  • Tests: tests/test_model_zoo/test_model_zoo.py — 37 tests
  • Key Classes:
  • PluginRegistry — Neuron model plugin system
  • VerilogGenerator — One-command Verilog emission
  • DocGenerator — Auto-documentation from plugins
  • LIFPlugin, IzhikevichPlugin, AdExPlugin, HodgkinHuxleyPlugin

Test Summary

Module Location Tests
safety_cert tests/test_safety_cert/ 456
asic_flow tests/test_asic_flow/ 100
fault_injection tests/test_fault_injection/ 22
uvm_gen tests/test_uvm_gen/ 72
hypervisor tests/test_hypervisor/ 78
twinsync tests/test_twinsync/ 72
spintronic tests/test_spintronic/ 66
chiplet tests/test_chiplet/ 172
memristor tests/test_memristor/ 70
analog_bridge tests/test_analog_bridge/ 27
bioware tests/test_bioware/ 200
meta_plasticity tests/test_meta_plasticity/ 72
evo_substrate tests/test_evo_substrate/ 160
federated tests/test_federated/ 93
bci_studio tests/test_bci/ 32
explainability tests/test_explainability/ 71
neuro_symbolic tests/test_neuro_symbolic/ 34
stochastic_doctor tests/test_stochastic_doctor/ 16
model_zoo tests/test_model_zoo/ 37
TOTAL 1,242

Running Tests

Bash
# Single module
python3 -m pytest tests/test_evo_substrate/ -q

# All integrated modules
python3 -m pytest tests/test_safety_cert tests/test_asic_flow tests/test_fault_injection \
  tests/test_uvm_gen tests/test_hypervisor tests/test_twinsync tests/test_spintronic \
  tests/test_chiplet tests/test_memristor tests/test_analog_bridge tests/test_bioware \
  tests/test_meta_plasticity tests/test_evo_substrate tests/test_federated tests/test_bci \
  tests/test_explainability tests/test_neuro_symbolic tests/test_stochastic_doctor \
  tests/test_model_zoo -q

Provenance

All modules were industrialized through a multi-pass gap analysis and hardening process. This document serves as the single source of truth for what was integrated, where it lives, and how it is tested.

Rust Crate Architecture

The Research Rust crates (tinysc_riscv, core_engine, autonomous_learning, neuro_symbolic) remain as standalone crates and are NOT merged into engine/. Rationale:

  1. engine/ is a single PyO3/maturin crate (sc_neurocore_engine) with cdylib + rlib output, custom build config, and bridge/ Python source path. Adding unrelated crates would break the maturin develop / pip install -e . workflow.
  2. The Research Rust crates are standalone research crates with no Python binding.
  3. When any of them matures to production, it should be added as a separate workspace member under a new top-level crates/ directory (not inside engine/).

Future Workspace Migration (when ready)

TOML
# SC-NEUROCORE/Cargo.toml (new workspace root)
[workspace]
members = [
    "engine",
    "crates/tinysc_riscv",
    "crates/neuro_symbolic",
]

This preserves the engine's build integrity while enabling shared dependency resolution.