On 2021-05-06 19:22, Marc Nieper-Wißkirchen wrote:
> Am Do., 6. Mai 2021 um 18:28 Uhr schrieb Amirouche
> <xxxxxx@hyper.dev>:
>
>> (lambda args
>> ;; TODO: use chibi's match
>
> NB: Use `case-lambda` when possible. It is optimized in a number of
> Schemes, while Chibi's match does not produce optimized code.

Thanks.

> More to the point:

> How does your code cooperate with call/cc called by the user or the
> exception system of Scheme (which includes exceptions and continuable
> exceptions).

It depends whether coop is implemented as a library or it is a builtin
ambiant coop.

If the Scheme has an ambient coop, there is no problems I can forsee.

When it is a library on top of a standard Scheme, so far I only worked
with an emulation of the ambiant coop. If the user capture a
continuation outside a flow, then I did not plan what can happen.

> Do you have to rewrite these things as well when you
> emulate delimited continuations with call/cc?

I do not understand the question. If the question is:

> Do you need to re-implement call/cc or exceptions or continuable
   exceptions to implement coop?

Then the answer is: no.

So far, the prototype is implemented as a Chez library, using call/cc,
call/1cc [1] and Chez POSIX threads procedures (a subset of SRFI-18).

[1] https://cisco.github.io/ChezScheme/csug9.5/control.html#./control:s9

I started to implement a Scheme that targets a micro subset of Scheme
where I re-implement call/cc with a trampoline plus CPS transformation.
The goal is to benchmark that against Chez's call/cc. Eventually try to
benchmark a program to be able to tell whether trampoline CPS
transformation is fast enough (and look at memory usage).

The problem I see with the current coop specification is that there is
no equivalent of parameters.

>
> Does your code respect proper tail recursion (see section 5.2 in [1])?
>

I did not read that article, yet.

Do tail calls recursions grow the stack with trampoline + CPS
transformation?

I have a working program that can compute factorial and fibonacci with
tail call recursions without crashing my computer.

> Marc
>
> -- [1]
> https://www.cambridge.org/core/journals/journal-of-functional-programming/article/abs/monadic-framework-for-delimited-continuations/D99D1394370DFA8EA8428D552B5D8E7E

--
Amirouche ~ https://hyper.dev