gc vs. no gc?

[Raoul Duke]

in the original paper it mentions being able to write (pre?)Scheme
code with a GC and then decide what parts to then take over for manual
memory management to get better performance in key places. has anybody
done this? does it mean starting with Scheme48 and then going back to
PreScheme or something?

i didn't see or missed the part where it explained how that would
really work since it say there is no gc at all in PreScheme.

thanks! (pre scheme is very exciting to a nerd like me.)

[Taylor R Campbell]

Date: Tue, 22 Jul 2014 11:03:18 -0700
From: Raoul Duke <***@gmail.com>

in the original paper it mentions being able to write (pre?)Scheme
code with a GC and then decide what parts to then take over for manual
memory management to get better performance in key places. has anybody
done this? does it mean starting with Scheme48 and then going back to
PreScheme or something?

I don't know what the paper is referring to, but PreScheme itself
doesn't have a GC, nor can it support one sensibly. You can write a
GC in PreScheme for a program like the Scheme48 VM, but all the
PreScheme code must take care not to hold onto Scheme objects in local
variables across a GC, just like the Scheme48 VM.

There's a partial reference manual for PreScheme, as it existed in
Scheme48 1.3, as part of

http://mumble.net/~campbell/darcs/s48-refman/

Hasn't been touched in almost a decade, and PreScheme may have changed
since then, so caveat lector.