Re: goops and memoization

[Neil Jerram <neil@ossau.uklinux.net>]

>>>>> "Mikael" == Mikael Djurfeldt <mdj@kvast.blakulla.net> writes:

    Mikael> Dirk Herrmann <dirk@sallust.ida.ing.tu-bs.de> writes:

    >> This conflicts with goops:  goops unmemoizes the function code, using
    >> 'procedure-source' (look into oop/goops/compile.scm).  This will re-create
    >> the original code, including all the symbols that refer to local
    >> variables.  This un-memoized code is then optimized in some way, and
    >> re-written into the closure object.  Then, if the closure is evaluated, it
    >> is not run through the memoizer again (since it is already a closure).
    Mikael>   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    Mikael> Just a note: This is something which holds *after* your
    Mikael> change, not previously. [...]  

    Mikael> It seems to me that what you need to do is to run the tail of the
    Mikael> cmethod (BODY-FORM1 ...) through your memoizer.  That should fix it.

Indeed.  If I understand correctly, what happens is that
scm_memoize_method in eval.c returns an unevaluated and unmemoized
lambda form, which the current evaluator handles by calling
scm_m_lambda, which performs the memoization stage.

The problem I presume is that your optimized evaluator doesn't handle
symbols in the CAR, because it shouldn't need to.

If this is correct, the fix is to pass `x' through your memoizer just
before `goto nontoplevel_begin;', like this:

	    apply_cmethod: /* inputs: z, arg1 */
	      {
		SCM formals = SCM_CMETHOD_FORMALS (z);
		env = EXTEND_ENV (formals, arg1, SCM_CMETHOD_ENV (z));
		x = SCM_CMETHOD_BODY (z);
                x = memoize (x);             /* ADDED */
		goto nontoplevel_begin;
	      }

A few notes:

- You could just call eval instead of memoizing and goto
  nontoplevel_begin, but that wouldn't be tail-recursive.  I wonder if
  tail recursion is important here?

- Is the memoization stage sufficient here?  What if the method
  definition came from a module with a syntax transformer in effect?
  If `procedure-source' returns post-transformation code, all is OK.
  If it doesn't, there are two possible problems:

  - The code returned by scm_memoize_method should be retransformed as
    well as rememoized.

  - `compile-method's manipulation of the source code might not work
    in the pre-transformation language.

- Quite a few places in the GOOPS code use local-eval.  Does
  local-eval still include memoization (and syntax transformation?) in
  your codebase?

I hope this helps; let me know if you would like me to dig or
experiment further.

        Neil



_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-devel