The patent that Jim Mulder cited: http://www.google.com/patents/US20090100243
contains very useful information. It begins with the text: <begin extract> A virtual storage technique is provided to manage a cell pool or a set of cell pools which can be used to satisfy variable-size storage requests. The algorithm uses no locks and relies on an atomic compare-and-swap instruction to serialize updates to the fields that can be simultaneously requested by multiple threads or processes. A free chain is used to manage cells which have already been obtained and freed, while there is an active extent that is used to hand out cells which have not previously been obtained. The algorithm is based on all cell pool extents being the same size, which allows the control information for the extent to be easily located on the extent boundary (e.g. at a 1 MB boundary). <end extract> The rationale for the use of the term 'pool' is now clear. It is that, while the cells allocated within a cell-pool extent may vary in size, cell-pool extents themselves are fixed--in fact 1 mibibyte--in size. This is legitimate developer-perspective terminology. It is less congenial from a user's perspective, but now that I understand its etiology I can live with it happily (which is just as well since I must do so whether happily or not). John Gilmore, Ashland, MA 01721 - USA ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
