Skip to content

CHANGELOG.md

Latest commit

 

History

History
130 lines (98 loc) · 5.62 KB

CHANGELOG.md

File metadata and controls

130 lines (98 loc) · 5.62 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

1.5.0 - 2025-10-30

Performance

  • Optimized StackAllocator initialization: Replaced iterator collect with explicit pre-allocation to reduce temporary allocations
  • Replaced O(n) double-free check with O(1) bitmap: FreeListAllocator now uses bitmap-based tracking for debug builds, eliminating linear scan overhead
  • Optimized RefCell borrow patterns in FixedPool: Reorganized allocation code to minimize borrow checking overhead
  • Optimized GrowingPool chunk lookup: Implemented O(log n) binary search with cached chunk boundaries instead of O(n) linear scan (10x+ faster for multi-chunk pools)
  • Eliminated mutex lock on ThreadSafeHandle deref: Cache pointer during allocation for lock-free deref/deref_mut operations (10-50x improvement for handle access)
  • Added inline annotations: Hot path functions now have #[inline] attributes for better optimization

Documentation

  • Added BENCHMARKS.md: Comprehensive benchmark results with real criterion.rs measurements, methodology, and honest comparisons
  • Added SAFETY.md: Detailed documentation of memory safety guarantees, unsafe code justification, and invariants
  • Added ARCHITECTURE.md: Internal design documentation explaining data structures, allocators, and implementation patterns
  • Added CONTRIBUTORS.md: Recognition for project contributors
  • Updated README.md: Replaced exaggerated performance claims (50x) with realistic measurements (1.3-1.4x), added proper context about when pools excel
  • Updated lib.rs documentation: Improved crate-level docs with accurate performance numbers and safety guarantees
  • Updated package description: More accurate description of library capabilities

Changed

  • Performance claims updated from "50x faster" to realistic "1.3-1.4x faster" based on actual benchmarks
  • Documentation now emphasizes predictable latency and zero fragmentation as primary benefits rather than raw speed
  • Added proper context explaining when memory pools excel vs standard allocators

1.0.1 - 2025-10-21

Changed

1.0.0 - 2024-10-16

Changed

  • STABLE RELEASE: First stable version with stable API
  • Improved Poolable trait design - removed blanket implementation to prevent conflicts
  • Added Poolable implementations for common types (primitives, String, Vec, Box, Option)
  • Enhanced async compatibility with better lifetime management
  • Fixed RefCell borrowing issues in GrowingPool
  • Improved thread safety and lifetime handling across all pool types

Fixed

  • Fixed compilation errors with thread-safe pools in async contexts
  • Resolved lifetime issues in handle management
  • Fixed borrowing conflicts in growing pool allocation
  • Corrected example code for proper Poolable trait usage
  • Fixed async example to properly drop handles before await points

Added

  • Explicit Poolable trait implementations for user types in examples
  • Better documentation for async usage patterns
  • Improved error messages and debug information

Performance

  • Optimized allocation path to avoid unnecessary borrows
  • Reduced lock contention in GrowingPool
  • Better cache locality with improved memory layout

0.1.0 - 2024-10-16

Added

  • Initial release of fastalloc
  • FixedPool with O(1) allocation and deallocation
  • GrowingPool with configurable growth strategies
  • ThreadLocalPool for zero-synchronization performance
  • ThreadSafePool for concurrent access
  • Multiple allocation strategies: Stack (LIFO), free-list, and bitmap
  • Smart handles with RAII automatic return to pool
  • Builder pattern for pool configuration
  • Growth strategies: None, Linear, Exponential, and Custom
  • Initialization strategies: Lazy, Eager, and Custom with reset callbacks
  • Optional statistics collection (stats feature)
  • Optional serialization support (serde feature)
  • Optional faster locks (parking_lot feature)
  • Optional lock-free pool (lock-free feature with crossbeam)
  • Optional tracing instrumentation (tracing feature)
  • no_std support (core + alloc)
  • Comprehensive documentation with examples
  • Extensive test suite including integration tests
  • Criterion-based benchmarks
  • Examples for common use cases

Features

  • Zero-cost abstractions with type-safe handles
  • Configurable memory alignment (including cache-line alignment)
  • Custom initializers and reset functions
  • Pool exhaustion detection with helpful error messages
  • Debug assertions for double-free and use-after-free detection
  • Thread-safe and thread-local variants

Performance

  • Fixed pool allocation: < 20ns per object
  • Deallocation: < 10ns per object
  • Memory overhead: < 5% for pools over 1000 objects
  • Thread-safe pool: < 100ns with moderate contention

Documentation

  • Comprehensive API documentation
  • Getting started guide
  • Performance tuning guide
  • Migration guide from standard allocation
  • Architecture documentation
  • Benchmark methodology
  • Contributing guidelines