«

Hi,

Does anyone know the reasoning or pros/cons for either (in a
multi-threaded C++ app):

- creating many sub-interpreters (Py_NewInterpreter) with a thread
state each

Or

- creating one interpreter with many thread states (PyThreadState_New)


When do I choose one approach over the other and why ?

Thanks

Alan

wrote:

> - creating many sub-interpreters (Py_NewInterpreter) with a thread
> state each
>
> Or
>
> - creating one interpreter with many thread states (PyThreadState_New)

My understanding is that using multiple interpeters isn't
really supported properly, despite there being apparent
support in the API. So I suggest using a single interpeter
with multiple threads.

--
Greg Ewing, Computer Science Dept,
University of Canterbury,
Christchurch, New Zealand
http://www.cosc.canterbury.ac.nz/~greg

Of course, with multiple threads in one process, it's important to know
about the dreaded global interpreter lock (GIL) --
http://docs.python.org/api/threads.html

With one instance of the interpreter, only one Python thread will be
running at a time, no matter how many CPUs you have in the system.


Greg Ewing wrote:
> wrote:
>
>> - creating many sub-interpreters (Py_NewInterpreter) with a thread
>> state each
>>
>> Or
>>
>> - creating one interpreter with many thread states (PyThreadState_New)
>
>
> My understanding is that using multiple interpeters isn't
> really supported properly, despite there being apparent
> support in the API. So I suggest using a single interpeter
> with multiple threads.
>

Greg Ewing wrote:
> wrote:
>
> > - creating many sub-interpreters (Py_NewInterpreter) with a thread
> > state each
> >
> > Or
> >
> > - creating one interpreter with many thread states (PyThreadState_New)
>
> My understanding is that using multiple interpeters isn't
> really supported properly, despite there being apparent
> support in the API. So I suggest using a single interpeter
> with multiple threads.

Huh. Mod_python uses multiple interpreters and within each
interpreter could be running multiple threads at one time
corresponding to different incoming HTTP requests.

It hasn't helped though that a bug was introduced in Python
2.3.5/4.0 which has been causing grief for some users creating
additional threads from within Python code itself.


http://sourceforge.net/tracker/index...70&atid=105470