"Joe "Nuke Me Xemu" Foster" <joe_at_bftsi0.UUCP> wrote in message
news:1070405687.459566_at_news-1.nethere.net...
> If the elements in a T-node are kept in order, use Vyssotsky's
> binary search. When designing the T-node layout, try to make
> it fit within the CPU level one data cache. This ought to be
> better than chasing binary tree node pointers all over slower
> RAM, which will cause the CPU to stall while RAM is copied to
> cache! While RAM is much faster than disk, level two cache is
> faster than RAM, and level one cache on the CPU is typically
> much faster than even the level two cache.
But then, by the same argument, T-Tree should be superior to B-Tree for
traditional databases as well, because it's just one level lower in the
memory hierarchy, right?
Received on Tue Dec 02 2003 - 17:58:46 CST
