Re: [Gcl-devel] intptr_t

[Matt Kaufmann]

Hi, Camm --

Excellent!  Thanks for the great work.

One thing -- the directions you gave for 2.6.8 and 2.6.9 are the same:

cvs -z9 -q -d:pserver:address@hidden co -d gcl-2.6.8 -r \
Version_2_6_8pre gcl

Should I simply change "8" to "9" twice for 2.6.9?

Also, you sent me a similar command that failed, but later corrected
it by replacing

  -d:pserver:address@hidden

with:

  -d:pserver:address@hidden:/sources/gcl

Tarballs would be perfect.  The command above (as amended) is close to
what you've given me previously, so I wonder if it could result in
different source bases depending on when it's executed.  I'd like to
have something static to point to.

Thanks --
-- Matt
   From: Camm Maguire <address@hidden>
   Cc: address@hidden
   Date: Thu, 06 Jun 2013 18:19:21 -0400

   Greetings!  What I'm thinking at the moment is to finalize a 2.6.8 as
   is, and a 2.6.9 with the listed improvements, within a few days,
   producing two final tarballs.  There is already a Version_2_6_9pre cvs
   tag which is just about ready.

   So, alas, since our schedules have skewed and you need to go to print:

   Right now the battle hardened 2.6.8 sources can be had here:

   cvs -z9 -q -d:pserver:address@hidden co -d gcl-2.6.8 -r \
   Version_2_6_8pre gcl

   and 2.6.9 here:

   cvs -z9 -q -d:pserver:address@hidden co -d gcl-2.6.8 -r \
   Version_2_6_8pre gcl

   Within a week or so I hope to have both in tarball form at the GNU
   website:

   http://www.gnu.org/software/gcl/

   The differences between 2.6.8 and 2.6.9 are in brief:

     * two word cons
     * word sized fixnum
     * immediate fixnums
     * dynamic maxpage

   As for the latter, gcl will now try to scale the maxpage limit
   dynamically given the memory environment on the executing machine.  The
   user can restrict this with #'si::set-log-maxpage-bound.  #'room also
   now accepts an optional argument to give more information.

   The thinking here is that these changes are too extreme for 2.6.8, yet
   logically go together.  I'd also like to have two relatively stable
   versions out there so people could test and determine whether or not all
   of these are actually improvements.  Two word conses in particular have
   some CPU overhead cost in exchange for 50% more cons space -- is this a
   good tradeoff?  

   2.6.9 is passing everything thus far.  All that really remains is to
   adjust the mac and windows builds, test on Debian, and perhaps put in
   configure switches to toggle the above.  sgc also needs a performance
   tweak to grow the page ranges properly.  You can see preliminary builds
   under /p/bin/gcl-2.6.9pre at ut.

   Take care,


   Matt Kaufmann <address@hidden> writes:

   > Hi, Camm --
   >
   > You say:
   >
   >>> 2.6.8 and below, fixnum == (signed-byte 32)
   >
   > So is the behavior below not something that would be in the "2.6.8"
   > that people might download?
   >
   >   dunnottar:~% /projects/acl2/lisps/gcl/2.6.8pre-2013-05-camm/b/usr/bin/gcl
   >   GCL (GNU Common Lisp)  2.6.8 CLtL1    May 24 2013 16:02:33
   >   Source License: LGPL(gcl,gmp), GPL(unexec,bfd,xgcl)
   >   Binary License:  GPL due to GPL'ed components: (XGCL READLINE UNEXEC)
   >   Modifications of this banner must retain notice of a compatible license
   >   Dedicated to the memory of W. Schelter
   >
   >   Use (help) to get some basic information on how to use GCL.
   >   Temporary directory for compiler files set to /tmp/
   >
   >   >most-positive-fixnum
   >
   >   9223372036854775807
   >
   >   >
   >
   > Thanks --
   > -- Matt
   >    From: Camm Maguire <address@hidden>
   >    Date: Wed, 05 Jun 2013 11:14:38 -0400
   >
   >    Greetings!
   >
   >    2.6.8 and below, fixnum == (signed-byte 32), so on 64bit machines,
   >    intptr_t will be (unsigned-byte 64), and a bignum.
   >
   >    2.6.9 and higher, intptr_t will be fixnum, i.e. a machine long, on all
   >    platforms. 
   >
   >    Of the patches I mentioned earlier for 2.6.9, 64bit fixnums is the
   >    simplest.  If there was great interest I could consider this in 2.6.8,
   >    but I'd prefer not to.
   >
   >    Take care, 
   >    -- 
   >    Camm Maguire                                        address@hidden
   >    
==========================================================================
   >    "The earth is but one country, and mankind its citizens."  --  
Baha'u'llah
   >
   >    _______________________________________________
   >    Gcl-devel mailing list
   >    address@hidden
   >    https://lists.gnu.org/mailman/listinfo/gcl-devel
   >
   >
   >
   >
   >

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