[gclist] Still no taker for a Meta-GC?
Fri, 9 Jul 1999 08:15:43 +0200
Dear GC experts,
I can think of _so_ many implementation techniques, that can be
combined with each other in exponentially so many different ways,
each time with different performance tradeoffs,
that I was wondering if rather than have to encode
every combination manually, the hard bug-prone way,
and not being able to tweak choices by fear a wreaking havoc,
there could not be a metaprogramming platform to specify
memory encodings with a domain-specific language,
and automatically obtain from such specification
a (provably?) _correct_ allocator and garbage-collector.
You could then tweak parameters and make dynamic choices with feedback
(possibly automatic such choices at major GC time,
using another suitable a metaprogram), to really optimize
the allocator for your application without having to pay
a large implementating and debugging price
for every (conceptually) small modification.
Since there are interactions between the memory allocator
and the mutator at large, such a specification would also
be part of the specification language for the target
of an accompanying code-generator, in the context
of a modular compiler or of a compiler-compiler
(something that could interest C-- implementers,
if the project still exists).
I have already suggested such thing in the past on the list.
Does the idea raise more enthusiasm now than then?
I'd _love_ to find a lab where work on such metaprograms.
And building such metaprograms could be an opportunity
for such forum as the present one to become a codevelopment forum
where experts share actual runnable code (assuming it's free software)
instead of mere implementable ideas.
Maybe I'm just being over idealistic
as of what computer science should be,
but I imagine that science would benefit considerably more
from such free-meta-software collaboration that it does from
But that's a completely different topic than GC,
so that if you're interested in discussions about this topic in general
rather than its particular instance as applied to GC,
I forward you to my article
"Metaprogramming and Free Availability of Sources"
or to the mailing-list firstname.lastname@example.org, also web-interfaced at
If only I could convince such great people
as those that compose this forum...
I understand that I may annoy you with my periodic requests
about free software and metaprogramming.
Since I pointed you to the article where I discuss this topic already,
I will try to resist annoying you again for a few more months/years,
unless of course the idea raises more interest now than before.
Followup-To: the right forum, which may be one (or many?)
of email@example.com or firstname.lastname@example.org or email@example.com...
[ "Faré" | VN: Ð£ng-Vû Bân | Join the TUNES project! http://www.tunes.org/ ]
[ FR: François-René Rideau | TUNES is a Useful, Nevertheless Expedient System ]
[ Reflection&Cybernethics | Project for a Free Reflective Computing System ]
Clothes make the man. Naked people have little or no influence on society.
-- Mark Twain