unofficial mirror of guile-devel@gnu.org 
 help / color / mirror / Atom feed
From: stefan <stefan.tampe@spray.se>
To: guile-devel@gnu.org
Subject: Re: unification
Date: Sat, 15 May 2010 01:06:31 +0200	[thread overview]
Message-ID: <201005150106.31333.stefan.tampe@spray.se> (raw)

Hi,

This will give some ideas of how I prefere to handle
unification. I add a few extra instructions. And I do not
expect you to swallow that. But at least it is a beutiful 
idea.

Consider an example of a unifing matcher e.g.

(def f ((a _) a) 
       ((c b) b))

currently this compiles to

,x f

Disassembly of #<procedure f (arg121)>:

   0    (assert-nargs-ee/locals 25)
   2    (unify-init)
   3    (br :L142)                      ;; -> 19
   7    (local-ref 0)                   ;; `arg121'
   9    (unify-cons)
  10    (local-set 1)                   ;; `a'
  12    (unify-cons)
  13    (drop)
  14    (make-eol)                      ;; ()
  15    (unify-eq)
  16    (local-ref 1)                   ;; `a'
  18    (return)
  19    (br :L143)                      ;; -> 36
  23    (local-ref 0)                   ;; `arg121'
  25    (unify-cons)
  26    (local-set 3)                   ;; `c'
  28    (unify-cons)
  29    (local-set 2)                   ;; `b'
  31    (make-eol)                      ;; ()
  32    (unify-eq)
  33    (local-ref 2)                   ;; `b'
  35    (return)
  36    (toplevel-ref 1)                ;; `error'
  38    (object-ref 2)                  ;; "no match in f"
  40    (tail-call 1)

2 initiates and stors the (br :L142) location if we need to jump to the 
next match, then executes 7  (this can be made better)

7.  pushes argument
9.  do a unifying cons. essentially (push cdr) (push car)
    but the unifying flavor. if not a cons go to L142 and initiate
10. set a to the pushed car and pop the car
12. do another cons sequence
13. pop
14. push '() and then check if eq? with the cdr
then the resulting code is executed.

if there is a failure we jump to 19, initiates the new next jump as :L143
and continue as before.

here is the code for unify-cons in c.


#define UNIFY_MODDED				\
  {						\
  if(!unify_modded)				\
    unify_modded = (int) SCM_UNPACK(gp_newframe());	\
  }

#define UNIFY_FAIL				\
  {						\
    if(!unify_modded)				\
      gp_unwind(FI_TO_SCM(unify_modded));	\
    ip = match_next;				\
    sp = stack_next;				\
    NEXT;					\
  }

VM_DEFINE_INSTRUCTION(123, unify_cons, "unify-cons",0,0,0)
{
  gp_lookup(GP_GETREF(*sp));
  sp--;
  if(GP_CONS(gp_ret.ref))
    {
      PUSH(GP_UNREF(gp_ret.id + 3));
      PUSH(GP_UNREF(gp_ret.id + 1));
      NEXT;
    }	  
  if(gp_ret.val == 0) //variable
    {
      SCM *ref;
      UNIFY_MODDED;
      ref = GP_GETREF(gp_mk_cons());
      gp_set_ref(gp_ret.id,GP_UNREF(ref));
      PUSH(GP_UNREF(ref + 3));
      PUSH(GP_UNREF(ref + 1));	    
      NEXT;
    }
  
  UNIFY_FAIL;
}

Have fun
/Stefan






             reply	other threads:[~2010-05-14 23:06 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-05-14 23:06 stefan [this message]
  -- strict thread matches above, loose matches on Subject: below --
2010-04-12 13:30 unification stefan
2010-04-13 11:55 ` unification Ludovic Courtès
2010-04-14 18:36   ` unification Andy Wingo
2010-05-14  9:15   ` unification stefan

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://www.gnu.org/software/guile/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=201005150106.31333.stefan.tampe@spray.se \
    --to=stefan.tampe@spray.se \
    --cc=guile-devel@gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).