Stackless Platform Independence?
frederic.giacometti at arakne.com
Mon Mar 4 22:31:08 CET 2002
Aahz Maruch wrote:
> In article <3C838150.97A86F52 at arakne.com>,
> Frederic Giacometti <frederic.giacometti at arakne.com> wrote:
> >Aahz Maruch wrote:
> >More precisely, the point of stackless is to dispense from OS threads
> >(and Python lock) overheads for running threaded Python bytecode.
> >However, OS threads are still required to take charge of calls to
> >blocking C function (usually from external C libraries) that one want
> >to run asynchronously to the PVM; there is no turn around to this.
> Sure there is: just use Python's standard threading capabilities and
> release the Global Interpreter Lock, just like you do now. I will grant
> that it's not the most convenient thing in the world, and maybe we'll
> want to fix this in the future, but I think that getting Stackless done
> and squared away is more important than resolving this in the cleanest
> possible fasion.
FYI, Python threads rely on the underlying OS thread switching + OS lock
synchronization API.... It's not because you don't see it that it is not used :))
Furthermore, as everyone in this thread knows by now, stackless is a pre-requisite
to Python microthreads. Thanks you for reminding it :))
More information about the Python-list