Time complexity of memory allocation

Question

What is the time complexity of dynamic memory allocation using new, malloc, etc.? I know very little about how memory allocators are implemented, but I assume the answer is

Answer 1

The time complexity for a heap allocator can be different on different systems, depending on what they might be optimizing for.

On desktop systems, the heap allocator probably uses a mixture of different strategies including caching recent allocations, lookaside lists for common allocation sizes, bins of memory chunks with certain size characteristics, etc. to try an keep allocation time down but also keep fragmentation manageable. See the notes for Doug Lea's malloc implementation for an overview of the various techniques that are used: http://g.oswego.edu/dl/html/malloc.html

For simpler systems, a strategy of first fit or best fit might be used, with the free blocks stored on a linked list (which would give a O(N) time with N being the number of free blocks). But a more sophisticated storage system such as an AVL tree might be used to be able to locate free blocks in O(log N) time (http://www.oocities.org/wkaras/heapmm/heapmm.html).

A realtime system might use an heap allocator like TLSF (Two-Level Segregate Fit), which has a O(1) allocation cost: http://www.gii.upv.es/tlsf/