Ben is a sophisticated benchmarking framework implemented in Rust that provides precise CPU cycle-level performance measurements, offering a powerful toolkit for analyzing and comparing Rust code performance.

Benchmark, query, and compare Rust performance.

• Measure function performance in CPU cycles.
• Query benchmarks with user-defined labels.
• Aggregate and compare function statistics.
• Display analysis in command-line tables.

See the mtr project for example benchmarks, queries and code.

• Precise CPU cycle measurement using low-level x86_64 processor instructions (RDTSC/RDTSCP)
• Memory fence operations for accurate timing measurements
• Parallel benchmark execution using thread pools
• Statistical analysis including median, average, min, and max calculations
• Query-based benchmark selection and comparison
• Command-line table visualization of results

• Custom traits and generics for flexible label management
• Zero-cost abstractions for performance measurements
• Safe abstractions over unsafe CPU instructions
• Thread-safe design with message passing
• Generic benchmark function handling with opaque function pointers
• Comprehensive error handling using anyhow

• Label-based benchmark organization and filtering
• Automated overhead calculation and compensation
• Customizable statistical analysis
• Support for manual timing control
• Built-in protection against compiler optimizations

• Direct CPU cycle measurement instead of system time for microsecond precision
• Parallel benchmark execution for efficient testing of large benchmark suites
• Memory fence instructions for accurate timing across CPU instruction reordering

• Query builder pattern for intuitive benchmark selection and comparison
• Type-safe label system using Rust's enum system
• Thread pool implementation for parallel benchmark execution
• Channel-based communication for benchmark results

• Safe abstractions over unsafe CPU instructions
• Comprehensive error handling for benchmark configuration
• Thread-safe design patterns
• Protection against compiler optimizations affecting benchmark accuracy

1. Study (Stdy): Main benchmark orchestrator
2. Registry Builder (RegBld): Benchmark function registration
3. Query Builder (QryBld): Benchmark selection and comparison
4. Statistical Analysis (Sta): Data processing and analysis
5. Table Visualization: Results presentation

• Pure Rust implementation
• x86_64 assembly instructions via Rust intrinsics
• Thread pools for parallel execution
• Message passing for thread communication
• Custom table formatting for result presentation

• Enables precise performance analysis of Rust code
• Supports data-driven optimization decisions
• Facilitates performance regression testing
• Provides insights into CPU-level code behavior

• Advanced Rust programming
• Low-level system programming
• Parallel computing implementation
• Performance optimization
• Safe abstraction design
• Technical documentation

.
├── Cargo.lock
├── Cargo.toml
├── examples
│   └── simple.rs
├── LICENSE
├── README.md
└── src
    ├── lib.rs
    ├── prv.1.rs
    ├── prv.2.rs
    ├── tbl.rs
    └── tst.rs

3 directories, 10 files