Module loading 

Model loading subsystem with WCET and cryptographic agility (spec §7.1) APR Loading Subsystem

Implements Toyota Way Heijunka (level loading) for model initialization. Prevents memory spikes in embedded environments with limited RAM.

Design Philosophy

Per Toyota Production System principles:

• Heijunka: Level resource demands during model initialization
• Jidoka: Built-in quality with verification at each layer
• Poka-yoke: Error-proofing via type-safe APIs

NASA Safety Compliance

Per NPR 7150.2D Section 3.6.1, all memory allocations are deterministic and bounded through pre-allocated buffer pools and streaming decompression.

References

• [Wilhelm et al. 2008] “The worst-case execution-time problem”
• [Liu & Layland 1973] Real-time scheduling theory
• ISO 26262 ASIL-B requirements for automotive ML inference

Re-exports

pub use cipher::CipherSuite;

pub use wcet::calculate_wcet;

pub use wcet::platforms;

pub use wcet::PlatformSpecs;

Modules

cipher

Cryptographic Agility (Post-Quantum Ready)

wcet

Worst-Case Execution Time (WCET) Calculator

Structs

BufferPool

Buffer pool for deterministic memory allocation

LoadConfig

Load configuration with Toyota Way Jidoka (quality built-in) enforcement

LoadResult

Result of a model load operation

Enums

Backend

Backend selection for Trueno operations

LoadingMode

Loading strategy selection based on deployment target

VerificationLevel

Verification level for model loading (NASA defense-in-depth)