From: Hoehle, Joerg-C. <Joe...@t-...> - 2003-10-01 16:14:57
|
Hi, >This is a lengthy discussion? one of your e-mails was quite as long as what I write from times to times :-) My opinion is that the functions dealing with Lisp arrays should continue to use uintL for the near future, or you're in for a really huge patch, and the functions that seek or maintain the current position use size_t, off_t or DWORD or whatever (CLISP had a tradition of using the exact types that the OS uses, e.g. DWORD). That's what I believe could work fastest (not execution time, I mean programmer time). >What is the size of a lisp >object, and is it (could it be? I hope so!) different on a 64 bit >machine? From what I remember from the early 64bit ports and -DWIDE configurations and don't know if that ever changed since, the use of bits was rather wasteful. An object was/is 64 bits, but with 32tag bits not fully used, and 32bits for e.g. fixnums. That means many unused bits. One might have expected something like >50 bits for fixnums and the rest for tagbits. See what MOST-POSITIVE-FIXNUM returns. We'll have a Sun here with 4GB RAM soon (for a client, I'm not going to compile CLISP there). I wonder what their memory map looks like, and whether a single process can take more than 4GB?? Regards, Jorg Hohle. |