25 Sep
2002
25 Sep
'02
4:15 p.m.
Markus Werle
Could You please tell us which one You used? The boost pool_allocators gave me exactly no difference in speed for my own project ...
I wrote one myself. I then was pointed to the optimized allocators that come with STLPort (based on work from SGI) and had similar results. My implementation stores deallocated chunks in a singly-linked list and tries to allocate from that list before asking a different allocator for the memory. This works quite well if you use one allocator per regex, since most chunks allocated when using one regex have the same size.