4.10 Summary

Some C++ idioms make heavy use of small objects allocated on the free store. This is because in C++, runtime polymorphism goes hand in hand with dynamic allocation and pointer/reference semantics. However, the default free store allocator (accessible through the global ::operator new and ::operator delete) is often optimized for allocating large objects, not small ones. This renders the default free store allocator unsuitable for allocating small objects because it is often slow and brings nonnegligible per-object memory overhead.

The solution is to rely on small-object allocators—specialized allocators that are tuned for dealing with small memory blocks (tens to hundreds of bytes). Small-object allocators use larger chunks of memory and organize them in ingenious ways to reduce space and time penalties. The C++ runtime support helps by providing the size of the block to be released. This information can be grabbed by simply using a less-known overloaded form of operator delete.

Is Loki's small-object allocator the fastest possible? Definitely not. Loki's allocator operates only within the confines of standard C++. As you saw throughout this chapter, issues such as alignment must be treated very conservatively, and being conservative spells being less than optimal. However, Loki is a reasonably fast, simple, and robust allocator that has the advantage of portability.