[EMAIL PROTECTED] wrote on 11/26/2007 10:52:30 AM: > How do buddy classloading and DynamicImport-Package compare with > regards to performance ? > I've had very bad experiences performance-wise with buddy classloading > and a global policy...
There is an existing bug about buddy classloading performance https://bugs.eclipse.org/bugs/show_bug.cgi?id=152006 This bug is about performance where there are a large number of bundles for the "registered" buddy policy. The "global" buddy policy can be expensive because we must search the "global" set of packages for the package where the requested class is from. We could optimize this to be faster but I really hope the "global" policy is rarely used. If you have a strong usecase for using the "global" policy and are seeing bad performance then please open a bug report so we can consider improving it. DynamicImport-Package does have a performance hit also, but it is only a one time hit when trying to establish a wire. Once the wire is established we no longer have to try and resolve the import on the next class request from the same package. This is one of the differences between dynamic import and buddy classloading. Dynamic imports are statically bound a package wire once they are established. In buddy loading a wire is not kept, the search for a buddy is done each class/resource load. Tom.
_______________________________________________ equinox-dev mailing list equinox-dev@eclipse.org https://dev.eclipse.org/mailman/listinfo/equinox-dev