Re: PLEASE COMMENT: faster elisp function calls

Glynn> Either you would need to relocate all of the pointers (and Glynn> Lisp code is primarily pointers) when you load the library, Isn't this exactly what all the gung-ho stuff is all about. Glynn> or every object which is referenced from the library would Glynn> have to have a specific address, which was known when the Glynn> Lisp was compiled. This is is about what 'a.out' style shared libraries where all about? May I propose a third option We introduce a new "internal" pointer type, that is basically just relative to the base adress of the packages. If the OS supports COW that would allow loading to be a simple mmap followed by a link of the external references. Furthermore it would increase the amount of shared data between XEmacsen. Of course this would also introduce a runtime penalty but I would like to see how big this is. Moreover the XEmacs could do some kind of just-in-time rewriting of the adress to the real pointer. Jan P.S. I have no experience with lisp implementation, but suppose we had lexical scoping wouldn't most of the references be stack relative anyway?