Re: faster serializer?

[Sascha Schumann <sascha@schumann.cx>] > On Fri, 3 Aug 2001, Brian Moon wrote: > > > I would love to see a faster serializer. I think serializing has a lot of > > upsides but the downside right now is performance. > > Microbenchmarking showed a two-fold speed decrease with the > patch I committed (serializing a couple of objects with a few > slots each). The serializer has to scan once through all > objects/slots, so we cannot expect it to become much faster > than it is right now. > > The deserializer code should be cleaned up, but I don't see > any significant potential on improving the speed of it. > > Anyway, there are only few applications which would > really gain from further refinement in this area. E.g., > real-life applications such as IRCG spend less than a percent > of their run-time in this part of the code. I known, that's why I was proposing that the people who would benefit enough from it could pool together a "reward" for a bleeding-fast implementation. I was thinking in the direction of doing memory allocation and initialization in fewer operations. - Stig -- Stig Sæther Bakken <ssb@fast.no> Fast Search & Transfer ASA, Trondheim, Norway