ghzcrlvct wrote:
each node being stored before the actual heap block a user allocates.
That's pretty specific to Doug Lea's malloc and its successor, ptmalloc which was based on it. Most allocators don't do that (like Hoard, jemalloc etc.).
The Linux kernel for example uses linked lists of
slabs, one node per partition (each partition stores allocations of the same quantum size, called slabs).
My allocator for example never mixes node data with application data, and it doesn't use linked lists at all (it has three different node types depending on quantum size).
Cheers,
bzt