Re: Q: unboxed values and polymorphism

[Hannes Haug <Hannes.Haug@Student.Uni-Tuebingen.de>]

Hi,

>>>>> "TT" == Tucker Taft <stt@henning.camb.inmet.com> writes:

    TT> Rather than tagged records, which allow for "unbounded"
    TT> polymorphism, you might want to simply use a variant record.
    TT> By using an appropriate record representation clause, you
    TT> should be able to keep your list cells down to 2 words.  You
    TT> might still find a use for unbounded polymorphism (i.e. tagged
    TT> records), when you point to larger objects which are not
    TT> simply composted of list cells.  However, if you are
    TT> essentially implementing a variant of Lisp, then I doubt if
    TT> tagged records will do much for you.  Tightly encoded variant
    TT> records seem more likely to be the answer.

But will this give me unboxed integers and type information on the stack ?
I'd try a private type that's actually an integer. I'd use half of this
integers (-2**30 ... 2**30-1) for my fixnums. Values outside this range
would be indices in arrays of bignums or list cells. So I'd have unboxed
integers and type information on the stack.

 - hannes