MPS issue job003324

TitleManual contains little about performance of interface
Assigned userGareth Rees
DescriptionRB pointed out in discussion that a developer trying to use the MPS will need to know more than just the functionality of each MPS function: they will need to know something about its performance. For example, is mps_addr_pool a slow function suitable for getting extra debugging information, or is it a super-fast function suitable for use on the application's critical path?
AnalysisAdd this information to the symbol reference (but only where we are prepared to support it, of course).

GDR: I'm not in a position at the moment to add this information myself, so I'll come back to it later.

DRJ's "Using the MPS Location Dependency Interface" [1] contained some performance information about the LD interface, but NB pointed out that it is wrong [2].

RB suggests, "If it's just O(count(threads)) then no problem. If you're registering O(heap size) roots then it's a problem! We should document that."
How foundinspection
Evidence[1] <>
[2] <>
[3] <>
Observed in1.110.0
Created byGareth Rees
Created on2012-10-18 16:48:49
Last modified byGareth Rees
Last modified on2014-06-08 12:23:24
History2012-10-18 GDR Created.