jemalloc/jemalloc
Jemalloc
jemalloc is a general purpose C memory allocator designed as a replacement for the standard library malloc and free functions. It is a multi-threaded allocation library that emphasizes fragmentation avoidance and scalable concurrency for high-performance applications.
The project reduces lock contention by using multiple independent memory arenas and thread-local cache layers. It minimizes latency by offloading the reclamation of unused memory pages to asynchronous background threads and utilizes huge page metadata storage to reduce translation lookaside buffer misses.
The system includes a heap profiling tool for analyzing allocation patterns and identifying memory leaks. It also provides custom memory management hooks that allow developers to override how the allocator requests and releases memory from the operating system.
Features
- Custom Memory Allocators - Provides a high-performance custom memory allocator designed to replace the standard C library malloc and free functions.
- Allocateurs de mémoire - Implements a general-purpose memory allocation library emphasizing fragmentation avoidance and scalable concurrency.
- High-Concurrency Memory Scaling - Optimizes how threads request and release memory to prevent performance bottlenecks in large-scale, high-concurrency software.
- High Performance Allocators - Provides a high-performance allocation implementation optimized for execution speed and reduced lock contention via multiple arenas.
- Arena-Based Memory Management - Implements an arena-based memory management strategy that binds threads to specific pools to reduce lock contention.
- Arena Allocators - Assigns threads to independent memory pools to reduce lock contention and improve data locality.
- Slab Allocators - Utilizes slab allocators to group objects of the same size, minimizing internal fragmentation and speeding up lookups.
- C Heap Allocators - Serves as a high-performance replacement for the standard C library malloc and free functions.
- Extent-Based Management - Uses extensible trees to track contiguous blocks of memory pages for efficient reclamation and fragmentation control.
- Multi-threaded Memory Management - Reduces lock contention and fragmentation when allocating memory across many CPU cores in high-performance environments.
- Thread Local Storage - Implements per-thread buffers to store frequently used memory chunks, eliminating the need for global locks.
- Metadata Address Mapping - Employs a multi-level radix tree structure to quickly map memory addresses to allocation metadata.
- Huge Page Optimizations - Utilizes transparent huge pages for internal metadata storage to minimize translation lookaside buffer misses.
- Low Latency Allocation - Optimizes allocation paths and offloads memory purging to background threads to ensure minimal latency.
- Heap Allocation Strategies - Ships a heap profiling tool that monitors allocation patterns and analyzes usage to detect leaks.
- Unused Memory Recovery - Offloads the recovery and return of unused memory pages to the operating system using asynchronous background threads.
- Custom Memory Hooks - Provides custom hooks to override default operating system memory request and release methods.
- Deferred Page Purging - Implements asynchronous background threads to return unused physical memory pages to the operating system and reduce latency.
- Memory Purging Tuning - Provides configuration options to tune the decay time and methods for purging unused memory back to the operating system.
- Memory Profilers - Includes a built-in tool for analyzing heap allocation patterns and identifying memory leaks.
- Memory Management - General-purpose allocator with concurrency support.
- Memory Allocation - General purpose malloc with fragmentation avoidance and concurrency support.
- Memory Management Libraries - Memory allocator emphasizing fragmentation avoidance.
