Re: feature request from r5rs

On Wed, Jun 11, 2003 at 02:11:50PM +0100, Matthias Radestock wrote:
> Narang wrote:
> >From r5rs: * Some implementations may implement ``implicit forcing,''
> >where the value of a promise is forced by primitive procedures like
> >cdr and +:
> >
> >(+ (delay (* 3 7)) 13)          ===>  34
> >
> >
> >Can this feature be added to sisc? thoughts.
> 
> It wouldn't be hard to add, with three caveats:
> 
> * preventing it from slowing down *everything* is difficult,
> 
> * synchronization is required in order to implement the correct 
> semantics in the presence of multiple threads,
> 
> * care needs to be taken to ensure the correct call/cc semantics is 
> implemented.
> 
> R5RS doesn't say anything explictly about the last point. However, it is 
> clear from the examples that capturing a k inside a delayed expression 
> and subsequently invoking it will
> 
> * (partially) re-evaluate the delayed expression and re-set the memoized 
> value,
> 
> * resume computation at the point where the force (explicit or implicit) 
> occured.

Fortunately our evaluation model covers correct call/cc semantics fairly 
easily, if done right.  The key is that any SISC uexp design should 
divide any sequence where a continuation should be captured into two 
uexps, the former being the computation, the latter being the effect.  
Hence the division into *Exp, *Eval.  So implicit forcing would be 
something to the effect of:

EvalExp -> PromiseSetEval

Where the EvalExp contains the body of the promise, and the 
PromiseSetEval thread-safely sets the memo box.

	Scott

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQE+5znbr9IW4v3mHtQRAi+tAJ9FmlHy2BcspKchhxXZkboe1QMsKgCfW1Ce
Uv6ohF4xQ6yC8szfVlkj3pQ=
=mHpU
-----END PGP SIGNATURE-----