On 01/13/2012 12:26 PM, Michael Meeks wrote:
Another thing that interests me is that parsing the configuration, despite all Stephan's good work in configmgr to speed it up, still takes real time and space: say 4% of our startup and a chunk of RAM. If we look at the sizes of our config using a simple:That makes me wonder whether either: a) we can/should split out the Labels tree until it is actually needed / used with some configmgr trick or b) use a different format / chunk of code for label descriptions
The decision to read in *all* data upfront in the new configmgr design was kind of a gamble. It significantly simplified the code and data layout, while appearing fast enough for contemporary desktop hardware to be acceptable. There is no easy "configmgr trick" to split out a chunk of data to be only read on demand. If we would need such a feature, it would be some work to add it. (And "a 3% CPU cycle / startup win" does not sound too exciting to me, being rather fine with the start up performance of LO from a user's perspective, at least on Linux and Mac OS X.)
However, I would prefer improving configmgr (if need be) over switching the label data to yet another format.
Stephan _______________________________________________ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
