Rust Engine API (sc_neurocore_engine)

The high-performance Rust engine provides SIMD-accelerated stochastic computing with SIMD-accelerated bitstream operations, IR compilation, and HDC support.

Browse the full Rust API documentation →

Key Modules

Module	Description
bitstream	Packed bitstream types and SIMD operations (AND, popcount, rotate)
encoder	LFSR-based stochastic encoders with decorrelated seeds
neuron	Fixed-point LIF neuron with Q8.8 arithmetic
layer	Dense layer pipeline with vectorised forward pass
ir	Intermediate representation for graph compilation
graph	Computational graph builder and verifier
attention	Stochastic attention mechanism
grad	Surrogate gradient training support
scpn	SCPN layer primitives (Petri net places/transitions)
simd	Platform-adaptive SIMD kernels (AVX2, SSE4.1, NEON, portable)
analysis	22 spike train analysis modules (see Rust Analysis Engine)
neurons	100+ neuron models — biophysical, maps, hardware, interneurons

Criterion Benchmarks → — all measured latencies for engine, neurons, and analysis.

Building from Source

Bash

cd engine
cargo build --release
cargo test
cargo doc --open

Python Bindings (PyO3)

The engine is exposed to Python via the local sc_neurocore_engine bridge. For source checkouts:

Bash

cd bridge
maturin develop --release

Python

import sc_neurocore_engine as engine

# Compile an IR graph
graph = engine.IRGraph()
graph.add_encode(0, 1024, 0xACE1)
graph.verify()
sv_code = graph.emit_sv()