home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ogicse!das-news.harvard.edu!cantaloupe.srv.cs.cmu.edu!crabapple.srv.cs.cmu.edu!andrew.cmu.edu!<UNAUTHENTICATED>+
- From: William.Lott@cs.cmu.edu
- Newsgroups: comp.lang.lisp
- Subject: Re: CMU-Lisp on 486 running Mach ?
- Message-ID: <0f2=FHG00jeiMrS8UM@cs.cmu.edu>
- Date: 16 Nov 92 22:58:59 GMT
- Article-I.D.: cs.0f2=FHG00jeiMrS8UM
- References: <1992Nov11.132604.19198@wavehh.hanse.de>
- Organization: Carnegie Mellon, Pittsburgh, PA
- Lines: 50
- In-Reply-To: <1992Nov11.132604.19198@wavehh.hanse.de>
-
- cracauer@wavehh.hanse.de (Martin Cracauer) writes:
- > I wonder, if one can bring up the CMU Common Lisp from the sources on
- > a PC running Mach.
- >
- > Anybody there, who knows ?
- >
- > An e-mail address of someone at CMU, who likes to respond, would be
- > good, too.
- >
- > Thanks
- > Martin
- > --
- > %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
- > Martin Cracauer, cracauer@wavehh.hanse.de
- > Waldstrasse 200, W-2000 Norderstedt, Germany
- > Tel.: +49 / 522 18 29, Fax.: +49 / 522 85 36
-
- Random queries about CMUCL can be sent to cmucl-bugs@cs.cmu.edu.
-
- A port to the 386/486 running MACH is currently in progress but is
- much more involved then a port to Yet Another Risc Machine would be
- because of the limited number of registers. The biggest problem is
- that the garbage collector assumes that the register file can be
- partitioned into two sets, one that only over holds valid lisp
- descriptors that can be identified by the garbage collector and one
- that holds raw binary values that would confuse the garbage collector.
- Given that the x86 only has 8 registers, artificially constraining
- their use would result in horrible performance if not a total
- implementation nightmare.
-
- We have figured out how to solve this problem, but it requires a
- rather significant change: the compiler is going to have to annotate
- the object code with information about what registers hold valid lisp
- objects (i.e. the liveness info) at various points in the code and
- then the garbage collector has to be taught that it can only GC at
- those points.
-
- We've been planning to move to a generational garbage collector, and
- it seemed like a good idea to merge these two tasks together given
- that they both involve bashing on the garbage collector. But other
- higher priority tasks keep pushing this one back. I plan on resuming
- active work on the 386/486 port Real Soon Now, and should have an
- alpha system Soon After That. In other words, don't expect anything
- before early next year, and that is even rather optimistic. (Of
- course, if someone were to donate money and/or machines to our
- project, that would of course cause us to re-prioritize the various
- tasks at hand. :-)
-
- -William Lott
- CMU Common Lisp Group
-