Is Stackless Python DEAD?
Donn Cave
donn at u.washington.edu
Mon Nov 5 16:35:13 EST 2001
Quoth "Mike C. Fletcher" <mcfletch at home.com>:
| I'm not sure I understand this argument. The current version is
| drop-in-compatible with the standard Python 2.0 DLL. You don't re-compile
| anything in extension modules, they just work the same as with the standard
| distribution.
|
| The patch is (apparently, haven't looked at it) intrusive in it's re-design
| of the core loop of the interpreter (it is re-writing some pretty basic
| mechanisms, after all), but for user-land (as distinct from
| interpreter-implementer-land) code, it's pretty much transparent in my
| experience.
|
| With that said, there are likely to be systems that don't work well under
| (for instance) micro-threading. Modules that need locks and expect
| micro-threads to look like real threads will be disappointed (1000s of them
| can be running in a single OS-level thread, so regular thread locks don't
| block the micro-threads). However, that's a problem only for those users
| actually using the Stackless-specific stuff (in essence if there are no
| micro-threads running, then the thread-assuming extensions work fine).
And that doesn't apply to continuations, they are compatible with OS
threads. That may be obvious, but FYI. In my experiment, I used lots
of C++ modules, some of which create OS threads that branch into the
interpreter and execute most of the program.
Donn Cave, donn at u.washington.edu
More information about the Python-list
mailing list