Re: Shrinking the C core

[Arthur Miller]

Po Lu <luangruo@yahoo.com> writes:

> Arthur Miller <arthur.miller@live.com> writes:
>
>> By the way, a single threaded shared state does not thread itself just
>> because it is transplanted over a multithreaded implementation :). Of
>> course. But getting access to such platform enables people to use
>> threads, while the implementation itself is tranformed, as well as buys
>> other benefits as discussed above.
>
> I'll reiterate this again: facilitating the execution of Lisp threads in
> parallel -- a more or less complete job -- is 5% of the work required
> for a concurrent Emacs.
>
> The tedious, insipid task of interlocking all that ``single threaded
> shared state'' follows and constitutes the bulk of that work, which
> someone must sign up for irrespective of the language it is written in.

Perhaps you are attacking the task from the wrong angle? I think that
sounds like a very tedius and error prone strategy. Why not refactor the
interpreter so you can instantiate it with an empty state in a separate
thread? I don't know; I am perhaps wrong, but isn't it more safe
strategy to isolate the state in their own threads instead of trying to
interloc bunch of stuff all over the place?