[Gcl-devel] Re: xgcl on Linux works

[Matt Kaufmann]

Hi --

Sorry, but I don't understand all the questions -- from a Lisp/Debian
perspective I'm a pretty naive user.  Here's my best effort:

1) Did you use the Debian package setup?

I don't think so.  I used whatever it was that Gordon Novak did for the P3
build.

2) Did you get readline support?

I think Gordon took that out.  By the way, I'm an emacs user and although I
don't know what readline support is, it's hard for me to imagine that it would
add any functionality that I'd want to use.  I encourage other ACL2 users
around me to use emacs as well.

3) Are you using gmp customized for a p3?  (I've never exercised the
        ISA extension code gmp will generate on such machines in a gcl
        program such as acl2 or maxima)

I don't know what "gmp" is or what it stands for.  Perhaps Kay knows.

4) Do you have an indication of gcl's performance
        strengths/weaknesses? 

Primarily, just the numbers on the test suite run that I CCed you on before,
which show that GCL 2.5.0 on a P3 is about a percent faster than Allegro 6.0
onj the same machine, for the ACL2 regression suite.

GCL:
14817.130u 121.280s 4:10:04.44 99.5%    0+0k 0+0io 5153829pf+0w

Allegro:
15527.850u 94.000s 4:21:11.23 99.6%     0+0k 0+0io 4881150pf+0w

I haven't done a CLISP or CMU Lisp run lately but in the past they've been
considerably slower, I think (though actually I don't think I've run CMU Lisp
on Linux).

In general Allegro compiles much more quickly than GCL, which makes the "win"
for GCL above all the more impressive, since there are perhaps 1500 files in
the test suite and nearly all are compiled (though many have only a few
definitions).

I think I've seen applications, however, where Allegro significantly
outperforms GCL.  Sorry I don't recall what they are, however.

-- Matt
   Cc: address@hidden, address@hidden, address@hidden,
      address@hidden
   From: Camm Maguire <address@hidden>
   Date: 28 Aug 2002 18:20:09 -0400

   Greetings!

   Great!  So...

   1) Did you use the Debian package setup?
   2) Did you get readline support?
   3) Are you using gmp customized for a p3?  (I've never exercised the
           ISA extension code gmp will generate on such machines in a gcl
           program such as acl2 or maxima)
   4) Do you have an indication of gcl's performance
           strengths/weaknesses? 

   Take care,

   Matt Kaufmann <address@hidden> writes:

   > Thanks, Gordon.  All ACL2 tests completed successfully on a P3 (craigievar)
   > using your build.  And, gcl-2.5.0-P3 beat Allegro 6.0 on craigievar.
   > 
   > GCL:
   > 14817.130u 121.280s 4:10:04.44 99.5%       0+0k 0+0io 5153829pf+0w
   > 
   > Allegro:
   > 15527.850u 94.000s 4:21:11.23 99.6%        0+0k 0+0io 4881150pf+0w
   > 
   > I also ran a small test suite on odin (which I guess is a P4) using
   > gcl-2.5.0-P3, without trouble.
   > 
   > Thanks (Kay and Camm too)!
   > 
   > -- Matt
   >    From: "Gordon Shaw Novak" <address@hidden>
   >    Date: Tue, 27 Aug 2002 18:09:40 -0500
   >    Cc: address@hidden, address@hidden
   > 
   >    I have made GCL-2.5.0-P3 which seems to run on both odin and loki.
   >    Give it a try.
   >    Cheers, Gordon
   > 
   > 

   -- 
   Camm Maguire                                         address@hidden
   ==========================================================================
   "The earth is but one country, and mankind its citizens."  --  Baha'u'llah