Re: [Algorithms] small part of large database
Brought to you by:
vexxed72
From: <mpo...@ed...> - 2001-09-10 09:28:04
|
> We need to uniquely identify up to 2^32 objects/entities. Only a few hundred > will ever be present in memory at once and I need fast acces to these by > their unique ID or knowledge that the referenced item is not present. The > cached object list is subject to very frequent and drastic change. > A lookup table would be too large and, as I said, access needs to be fast. Do you tried to bench with simple STL containers ? Perhaps a map could be simple enough to perform what you are looking for ? You write that you have only few hundred of ID's. Statisticaly, are they using only a part of the 2^32 potential ID's, or are they covering all the range of values ? Some algorithms will be efficient only if the statistic repartition of values is homogeneous. Mickael Pointier PS: Looks like I'm lacking english vocabulary in the "statistics" field :'( |