At 9:00 AM -0500 12/31/02, Jim Cromie wrote:
Lisp systems have done this on and off for decades. (Emacs did it for ages with the high bits, which is why there was a 4M buffer size limit for so long) It works best if you have control over the hardware, because then you could rope in MMU and page fault support to do Interesting Things.Dan Sugalski wrote:You're looking to encode data in the low X bits (where X is 2 or 4 for 32 or 64 bit machines), right?
yes- precisely!It's a clever thing to do, but one I admit I mistrust. Either it means we have to check every pointer that might be encoded before we use it or wedge into the system SIGBUS handler, which is non-trivial in many places. It's also rather a portability problem. (And potential security problem, alas)thats what I thought - but had to ask on the off-chance it could benefit.
A cool idea, but I don't think we can do it.
IIRC - I saw it once as a feature of some old Motorola ?? DMA controller - I think it was used to signal some scatter/gather ops
Believe me, I'd be *thrilled* to be able to leverage this for GC purposes, amongst other things. It's really platform-dependent, though, and requires kernel level access to leverage. (So it's really sensitive to dodgy code, too) Unfortunately rules it out for us.
--
Dan
--------------------------------------"it's like this"-------------------
Dan Sugalski even samurai
[EMAIL PROTECTED] have teddy bears and even
teddy bears get drunk
