[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: So, what the heck is a continuation anyway?

At 09:58 PM 12/10/01, Tony Kimball wrote:
>: Unless you heap allocate all values, copying the stack naively
>: will be a problem.
>This suggests a question, which may be of interest to many list
>members: Is there some good academic work on how to use
>garbage-collected heap activation frames without an enormous
>performance penalty on commodity CPUs?

I think Appel's "Compiling With Continuations" says that SML/NJ
does this.  I could be mis-remembering, I haven't looked at that
book for at list 6 ot 7 years.

A friend of mine who worked on a GC for a commercial ML implementation
claimed that with heap activation frame, you ended up consing like
crazy, but the consing behavior was so predictable, that the GC could
be made to work immensely rapidly, and he had numbers to prove that
his claim was correct for that ML implementation.  (This was for
Harlequin's ML implementation, if anybody ever used it.  It might have
had nearly one customer at the peak of its popularity.)