Chapter 18 -- Heaps

[Previous] [Next]

Chapter 18

The third and last mechanism for manipulating memory is the use of heaps. Heaps are great for allocating lots of small blocks of data. For example, linked lists and trees are best managed using heaps rather than the virtual memory techniques discussed in Chapter 15 or the memory-mapped file techniques discussed in Chapter 17. The advantage of heaps is that they allow you to ignore all the allocation granularity and page boundary stuff and concentrate on the task at hand. The disadvantage of heaps is that allocating and freeing memory blocks is slower than the other mechanisms and you lose the direct control over the committing and decommitting of physical storage.

Internally, a heap is a region of reserved address space. Initially, most of the pages within the reserved region are not committed with physical storage. As you make more allocations from the heap, the heap manager commits more physical storage to the heap. This physical storage is always allocated from the system's paging file. As you free blocks within a heap, the heap manager decommits the physical storage.

Microsoft does not document the exact rules that the heap follows for committing and decommitting storage—and the rules are different for Microsoft Windows 98 and Windows 2000. I can tell you this: Windows 98 is more concerned with memory usage, so it decommits heap storage as soon as possible. Windows 2000 is more concerned with speed, so it tends to hold on to physical storage longer and returns it to the paging file only after the pages have gone unused for some period of time. Microsoft is constantly performing stress tests and running different scenarios to determine the rules that work best most of the time. As applications and the hardware that runs them change, these rules will change. If this knowledge is critical to your application, don't use heaps. Instead, use the virtual memory functions (that is, VirtualAlloc and VirtualFree) so you can control these rules yourself.