Module compute 

Compute infrastructure integration (trueno 0.8.7+) Compute Infrastructure Integration (trueno 0.8.8+)

This module provides ML-specific wrappers around trueno’s compute primitives and simulation testing infrastructure.

Features

• Backend Selection: Automatic CPU/GPU dispatch based on data size
• Jidoka Guards: NaN/Inf detection for training stability
• Reproducibility: Deterministic RNG for reproducible experiments
• Stress Testing: Performance validation infrastructure

Toyota Way Principles

• Jidoka: Built-in quality - stop on defect (NaN/Inf detection)
• Poka-Yoke: Mistake-proofing via type-safe backend selection
• Heijunka: Leveled testing across compute backends

Example

use aprender::compute::{TrainingGuard, select_backend, BackendCategory};

// Auto-select backend based on data size (100k+ → GPU, 1k+ → SimdParallel)
let data_size = 50_000;
let gpu_available = false;
let backend = select_backend(data_size, gpu_available);
assert_eq!(backend, BackendCategory::SimdParallel);

// Create training guard for NaN/Inf detection
let guard = TrainingGuard::new("gradient_update");

// Check weights after update (detects NaN/Inf)
let weights = vec![1.0f32, 2.0, 3.0];
assert!(guard.check_weights(&weights).is_ok());

Structs

BackendSelector

Poka-Yoke: Type-safe backend selection

BackendTolerance

Backend-specific tolerance configuration

DivergenceGuard

Guard for detecting cross-backend numerical divergence

ExperimentSeed

Seed management for reproducible ML experiments

JidokaGuard

Jidoka guard for simulation tests

Matrix

A 2D matrix with row-major storage

StressTestConfig

Stress test configuration for trueno operations

StressThresholds

Performance thresholds for stress testing

TrainingGuard

ML-specific training guard for numerical stability

Vector

High-performance vector with multi-backend support

Enums

Backend

Backend execution target

BackendCategory

Backend category for selection result

JidokaAction

Jidoka action on condition trigger

JidokaCondition

Jidoka condition that triggers stop

JidokaError

Jidoka error types

Functions

select_backend

Select optimal backend for ML operation based on data size

should_use_gpu

Check if GPU would be beneficial for this operation size

should_use_parallel

Check if parallel execution would be beneficial