Re: Optimizing synchronization objects

[Alexandre Julliard]

Daniel Santos  writes:
> That is actually a wonderful idea! The struct ntdll_object &
> associated rbtree and list added to om.c in the patch set is exactly
> for this process-level cache. Rather the object is inter-process or
> not can just be data in that cache object. This would seem to add the
> most for the least complexity and exposure to security and corruption
> issues. This approach should fix most performance issues and still be
> very simple! :) Thanks!

If you think that it's very simple then probably you are missing
something ;-)

I think it's doable, at least for some object types, but it's far from
simple, there's still a lot of complexity involved. Consider for
instance a DuplicateHandle call from a different process.