Most of us use malloc() without thinking about what happens underneath.

I decided to implement my own memory allocator in C to understand it better. This wasn’t for production use, just to learn how allocation, fragmentation, and concurrency actually behave in practice.

I also benchmarked it against glibc’s malloc to see where it stands.

Implementation Overview

My allocator currently includes:

Thread-local cache

Free lists (bins) for different size ranges

Direct mmap for larger allocations

A custom realloc() implementation

Benchmark Results

glibc malloc

Single-threaded:

alloc + free (1M iterations): ~26 ms

batch alloc/free:3.40ms/0.95ms

mixed sizes: ~2.5 ms

Multi-threaded:

8 threads: ~57 ms

My Allocator

Single-threaded:

alloc + free (1M iterations): ~83 ms

batch alloc/free:1.46ms/0.50ms (faster than glibc)

mixed sizes: ~126 ms

Multi-threaded:

8 threads: ~791 ms

What Worked

Batch allocation and free operations were faster than glibc.

This likely comes from:

simpler logic in the fast…