Module profiling_optimizer 

Profiling-guided optimization for adaptive performance tuning.

This module provides runtime profiling and adaptive optimization:

• Profile collection: Gather execution statistics during runtime
• Hotspot detection: Identify performance bottlenecks
• Adaptive optimization: Adjust strategy based on observed behavior
• A/B testing: Compare optimization strategies
• Auto-tuning: Automatically select best configurations

Example

use tensorlogic_infer::{ProfilingOptimizer, OptimizationGoal, TuningConfig};

// Create profiling-guided optimizer
let mut optimizer = ProfilingOptimizer::new()
    .with_goal(OptimizationGoal::MinimizeLatency)
    .with_tuning_enabled(true);

// Execute with profiling
for batch in dataset {
    let result = optimizer.execute_and_profile(&graph, &batch)?;

    // Optimizer automatically adapts based on observed performance
    if optimizer.should_reoptimize() {
        optimizer.apply_optimizations(&graph)?;
    }
}

// Get optimization report
let report = optimizer.generate_report();
println!("Speedup: {:.2}x", report.speedup);

Structs

ExecutionProfile

Execution profile for a single run.

Hotspot

Hotspot in the computation graph.

OptimizationReport

Optimization report.

OptimizationStrategy

Optimization strategy configuration.

ProfilingOptimizer

Profiling-guided optimizer.

TuningConfig

Tuning configuration.

Enums

OptimizationGoal

Optimization goal.

ProfilingOptimizerError

Profiling-guided optimization errors.