Re: Languages don't matter. A mathematical refutation

[Paul Rubin <no.email@nospam.invalid>]

"Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> writes:
>> Dennis said specifically, his program needs determinstic time.  Some
>> programs are like that.
> He could run GC task at the lowest priority level.

I got the impression that there was no room for timing variation at all
in the program, and no idle time during which a GC could run.
E.g. something like a controller inside a jet engine, that has to do
something synchronized with the turbine spinning and can never be off by
even a microsecond.

> **  Most compilers use arena pools for data structures. It would be insane
> to deploy GC there, IMO.

Some compilers written in old, non-gc languages use arenas as a simple
way to get some of gc's benefits.  But compiler-writing is the
archetypal application of languages like Haskell, which rely on GC.
CompCert (compcert.inria.fr) could be seen from a certain perspective as
the world's only truly modern compiler.  It's written mostly in Coq,
which could be thought of as an Ocaml dialect with a fancier type
system.  Coq "compiles" by extracting Ocaml code from the Coq program
after type checking, so you can compile and run the extracted Ocaml
code.  The Coq code is far removed from anything like memory allocation
so it has to rely completely on gc.