My personal approach to arena allocation. Read more about the design and motivation of this allocator here: https://ishandutta2007.github.io/blog/arena
The code is implemented in C as an STB-style single-header library. The header is portable and can be used on its own. But platform-specific source files are also available to “configure” the library to use the system page allocator. This separation breaks the rule of header-only libraries, but it allows for easier addition of target platforms.
Using the library as a single-header header library is the same as any other
library of the same type: select exactly one C source file that instantiates
the code by defining ARENA_IMPLEMENTATION.
#define ARENA_IMPLEMENTATION
#include "arena.h"
The platform-specific source files act as the instantiation source files. So if
you use one of those, don’t define ARENA_IMPLEMENTATION anywhere else.
To add a new platform, redefine these two macros to call your platform’s allocator.
#define ARENA_BACKEND_ALLOC(size) ...
#define ARENA_BACKEND_DEALLOC(ptr, size) ...
#define ARENA_IMPLEMENTATION
#include "arena.h"
You can also define ARENA_BACKEND_RESIZE if your platform supports efficient
reallocation, but it is not required, as the library can automatically
implement that with ARENA_BACKEND_ALLOC and ARENA_BACKEND_DEALLOC.
Some macros that you can redefine to configure the library:
| Macro | Usage |
|---|---|
ARENA_MEMCPY |
Redefine if your platform can’t use libc’s memcpy |
ARENA_ALIGNMENT |
Customize the memory alignment. Default is 16 |
ARENA_MAX_SIZE |
The maximum size of each chunk, should be page-aligned |
This project is licensed under the MIT License.