Avoid prev/next pointers on a really big list?
I want a
sizeof(Node)=24.
Standard allocator works with one new per element, and prev_element, next_element pointers per element.
I don't know, for every allocation, if new implementantion waste memory bytes for internal reasons, but 2 pointers per element is a big waste of memory.
One more restriction is that Node* will be referenced from other (many) structures (so a vector which reallocates is a little bit mess).
Because I don't want to delete, 'gui-deleted' elements for UNDO reasons, I want to create my allocator with bigger allocated blocks:
I think something like a modified list of fixed length arrays of Node. Forward and backward pointers of list will only apply on fixed length arrays. Not on elements inside them.
Your opinion?
std::list<Node> with 10(at least) to 4000000000(at ambitious) elements.sizeof(Node)=24.
Standard allocator works with one new per element, and prev_element, next_element pointers per element.
I don't know, for every allocation, if new implementantion waste memory bytes for internal reasons, but 2 pointers per element is a big waste of memory.
One more restriction is that Node* will be referenced from other (many) structures (so a vector which reallocates is a little bit mess).
Because I don't want to delete, 'gui-deleted' elements for UNDO reasons, I want to create my allocator with bigger allocated blocks:
I think something like a modified list of fixed length arrays of Node. Forward and backward pointers of list will only apply on fixed length arrays. Not on elements inside them.
Your opinion?
| I want a std::list<Node> with 10(at least) to 4000000000(at ambitious) elements. sizeof(Node)=24. |
4 billion bytes is about 3.7 gigs. Times that by the size of your node, and you'll see that you don't have enough RAM to do that.
std::list takes care of the Node part for you, so all you need to give it is the object itself. You could also have the std::list store pointers to your object to reduce the amount of the stack that you use.
Maybe a disc based random access db such as Berkeley DB
http://en.wikipedia.org/wiki/Berkeley_DB
or sqlite
http://www.sqlite.org/
might be the way to go?
http://en.wikipedia.org/wiki/Berkeley_DB
or sqlite
http://www.sqlite.org/
might be the way to go?
Topic archived. No new replies allowed.