I have a working version of Qi under CMU Lisp 19c. The only catch is
that the (save) command for CMU I'm using does not seem to work. Its
hard to know whether this is because of some local feature of the
installation or because the command needs changing.
Otherwise the release seems to work fine. Here is the definition of
save; it is of course, a platform-specific command.
(DEFUN save ()
(SETQ *history* NIL)
(LET ((Imp (LISP-IMPLEMENTATION-TYPE)))
(COND ((EQUAL Imp "CLISP")) (SAVEINITMEM))
((EQUAL Imp "CMU Common Lisp")
(EXTENSIONS::SAVE-LISP "Qi" :INIT-FUNCTION
'qi::qi :PRINT-HERALD NIL))
(T (ERROR "Unknown platform to Qi: ~A" Imp)))))
Any suggestions from CMU Lispers welcome.
The CMU release shows a 400% improvement in performance over the CLisp
version and will appear on the web next month. It is available by
request for people to try on their machines. Just email me if you want
to try it.
Mark
> The only catch is that the (save) command for CMU I'm using does not seem to work.
If you want to know the system returns something like KERNEL SIMPLE
ERROR when save is invoked - hence my wondering if this is an
installation problem on this machine.
Mark
Mark Tarver wrote:
> > The only catch is that the (save) command for CMU I'm using does not seem to work.
>
> If you want to know the system returns something like KERNEL SIMPLE
> ERROR when save is invoked - hence my wondering if this is an
> installation problem on this machine.
>
> Mark
Sorry ignore this. Its fixed. The command was right but the bracket
was in the wrong place.
The prerelease now installs an image. The system is available on
request - email to me.
Mark Tarver
··········@ukonline.co.uk
André Thieme wrote:
> Great news.
> Can you already foresee how hard it could be to port it to
> sbcl as well?
>
>
> André
> --
Well I did a successful run late last night and I installed Qi under
CMU with telephone communication with Carl Shapiro in California.
(Keeping typical hackers hours of 3.00AM hence missing the bracket
above; Carl is on a more civilised schedule 8 hours behind me).
I can't think that SBCL can be very difficult because as I understand
it, its a fork from CMU. The SBCL for Windows may be a problem if it
is not stable. The performance figures under CMU are very good and I've
put together a set of timings for the web. I should predict that the
timings for SBCL would be very similar to CMU.
The main obstacle is the atrocious Linux installation I'm using with
Firefox disabled and much functionality not there. Hence part of the
reason for Carl and I interacting is to determine if the problems are
arising from my platform. At SRI he has far better support.
I'm going to fix this in January but for various reasons, including the
failure of my backup machine, I'll soldier on with this version for
another week or so until my backup arrives.
Preliminary results are at
http://www.lambdassociates.org/Download/benchmarks.htm. I'm interested
in comparing Qi under CMU with NJML, but that will have to wait for the
moment.
Mark