Struct OptimizationPipeline 

pub struct OptimizationPipeline { /* private fields */ }

Multi-pass optimization pipeline for TLExpr expressions.

The pipeline applies 7 optimization passes in sequence, iterating until a fixed point is reached or the maximum number of iterations is hit.

Pass Order

1. Negation optimization: Applies De Morgan’s laws and eliminates double negations. This exposes more opportunities for subsequent passes.

2. Constant folding: Evaluates constant expressions at compile time. This creates simpler expressions with fewer operations.

3. Algebraic simplification: Applies mathematical identities like x + 0 = x, x * 1 = x, etc. This can create new constants for folding.

4. Strength reduction: Replaces expensive operations with cheaper equivalents (e.g., x^2 → x*x, exp(log(x)) → x).

5. Distributivity: Factors common subexpressions to reduce redundant computation (e.g., ab + ac → a*(b+c)).

6. Quantifier optimization: Hoists loop-invariant expressions out of quantifiers (e.g., ∃x.(a + p(x)) → a + ∃x.p(x)).

7. Dead code elimination: Removes unreachable code and eliminates branches with constant conditions (e.g., if true then A else B → A).

Fixed Point Detection

The pipeline tracks whether each pass makes changes. If an entire iteration produces no changes (i.e., the expression is unchanged), a fixed point has been reached and optimization stops early.

Examples

use tensorlogic_compiler::optimize::{OptimizationPipeline, PipelineConfig};
use tensorlogic_ir::{TLExpr, Term};

// Default pipeline
let pipeline = OptimizationPipeline::new();

// Custom configuration
let config = PipelineConfig::default()
    .with_max_iterations(5)
    .with_constant_folding(true);
let pipeline = OptimizationPipeline::with_config(config);

// Optimize an expression
let expr = TLExpr::add(
    TLExpr::mul(TLExpr::Constant(2.0), TLExpr::Constant(3.0)),
    TLExpr::Constant(0.0)
);
let (optimized, stats) = pipeline.optimize(&expr);

Implementations

impl OptimizationPipeline

pub fn new() -> Self

Create a new optimization pipeline with default configuration.

pub fn with_config(config: PipelineConfig) -> Self

Create a new optimization pipeline with custom configuration.

pub fn optimize(&self, expr: &TLExpr) -> (TLExpr, PipelineStats)

Optimize an expression using the configured pipeline.

Returns the optimized expression and statistics about the optimizations applied.

pub fn config(&self) -> &PipelineConfig

Get the current configuration.

Trait Implementations

impl Default for OptimizationPipeline

fn default() -> Self

Returns the “default value” for a type.