Skip to content

Fast memory allocation routines for fixed size memory blocks

License

MIT, Unknown licenses found

Licenses found

MIT
LICENSE
Unknown
COPYING
Notifications You must be signed in to change notification settings

ronflima/zapalloc

Repository files navigation

zapalloc

Build Status License

The irresponsible memory allocator

Fast memory allocation routines for fixed size memory blocks. This small library is intended to provide fast allocation and memory reusage when fixed size memory blocks are needed. This assumption makes this library not suitable for general usage and is specific for several small scenarios where memory is needed in blocks of the same size.

You can create several contexts, however, each one taking care of a block of a different size.

This is not a memory debugger, but an attempt to optimize memory allocations by given the opportunity to alloc large amounts of memory and manage usage for you.

Fast or Crash

This library was created with this in mind: be fast or crash. There is no memory debugging, no memory protection, no thread protection, no nothing. So, if your software invade memory, it will crash. So, be careful.

Simple, stupid

This library was built over STDC malloc/calloc/realloc functions at this version. In the future, it will replace, internally, STDC functions in favor of system calls that can be faster. It was designed to minimize system calls when returning memory to the user. Algorithms are really simple and are designed for speed.

Concepts

All memory is attached to a context. A context manages a certain amount of memory which is defined during its initialization. If you request more memory than it was initialized, the library will increase the internal memory in order to return to you more memory, but using its initial definitions.

During its initialization, you must provide the number of blocks and the size of each block. ZapAlloc will initialize your context with so much memory already allocated inside. For instance, let's say you asked for 1024 blocks of 2048 bytes each. During initialization, ZapAlloc will create a poll of 1024 blocks, with 2048 bytes each. Each call to alloc function will return to you a block containing 2048 bytes. When you request the 1025th block, ZapAlloc will create a new poll containing more 1024 blocks of 2048 bytes each and return a 2048 bytes block to you.

The ideia behind the scenes is to:

  • centralize all memory management on a single manageable place (the context);
  • request a lot of memory at once, but fewer times to the system, reducing the quantity of system calls, since system calls are too expensive.

Why don't rely on malloc itself?

The STDC malloc function is great for general memory allocation. Some implementations do change the process's core size in order to return memory to the system, which is great for general purpose. However, if you want performance, it is a great idea avoiding to call the system too many times. Each system call will cost you a lot of processing time, since your process will be interrupted by the operating system in order to have it's core resized.

So, zapalloc try to avoid those many system calls by allocating a lot of memory at once and then managing the memory blocks in a simple fashion, so you can have a great performance at the end.

License

This project is released under the MIT license. See LICENSE for details.

About

Fast memory allocation routines for fixed size memory blocks

Topics

Resources

License

MIT, Unknown licenses found

Licenses found

MIT
LICENSE
Unknown
COPYING

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published