Module linear 

Linear type system for resource management in TensorLogic.

This module implements linear types (also known as affine types), where values must be used exactly once. This is crucial for:

• Memory management: Ensuring tensors are properly deallocated
• In-place operations: Tracking when tensors can be mutated safely
• Resource tracking: Managing GPU memory, file handles, etc.
• Side effect control: Ensuring operations execute in the correct order

Examples

use tensorlogic_ir::linear::{LinearType, Multiplicity, LinearContext};

// Linear type: must be used exactly once
let tensor_handle = LinearType::linear("TensorHandle");

// Unrestricted type: can be used multiple times
let int_type = LinearType::unrestricted("Int");

// Check multiplicity constraints
let mut ctx = LinearContext::new();
ctx.bind("x", tensor_handle);
assert!(ctx.is_linear("x"));

Multiplicity System

• Linear (1): Must be used exactly once
• Affine (0..1): Must be used at most once
• Relevant (1..): Must be used at least once
• Unrestricted (0..): Can be used any number of times

Structs

LinearContext

Linear typing context for tracking variable usage.

LinearResource

Linear resource with capabilities.

LinearType

Linear type: type with multiplicity constraints.

LinearityChecker

Linearity checker for expressions.

Usage

Usage tracking for linear variables.

Enums

Capability

Capability: describes what operations are allowed on a linear resource.

Multiplicity

Multiplicity: how many times a value can be used.