In C++, what is a memory arena?

This article is from: https://www.quora.com/In-C++-what-is-a-memory-arena
The original author holds the copyright.

---

A memory arena is simply a large, contiguous chunk of memory that is allocated once and then used to manage memory manually by handing out smaller chunks of that memory.

The point of all this is usually to improve performance when you know how you will be using memory. For example, you may be allocating and deallocating blocks only of a fixed size, in which case you can write very efficient reclamation algorithms. Often when using memory arenas you can also free the entire arena at once when you are finished with it, instead of requiring individual calls to library free functions for each small chunk. When you have many small chunks of memory this can reduce overhead significantly.

When using C++, it should be noted that if you are allocating objects from an arena and you want fast deallocation of the entire arena you should consider if you need destructors to be called. If so, this will require calling them manually on all the objects within the arena. Along these lines, if you can place only POD types in your arena you will be able to reduce allocation/deallocation overhead to a minimum.