Re: [Python-Dev] Sub-interpreters: importing numpy causes hang
Reading through that post, I think I have everything covered but this here:
- The third and final scenario, and the one where the extended GIL
state functions for Ensure is still required, is where code doesn't
have the GIL as yet and wants to make a call into sub interpreter
rather than the main interpreter, where it already has a pointer to
the sub interpreter and nothing more. In this case the new
PyGILState_EnsureEx() function is used, with the sub interpreter being
passed as argument.
If I understand it correctly, it means the following in practice:
Whenever I or a third-party library start a new thread, we need to
query what interpreter we are running at the moment (in the thread
that is starting the new thread) and pass that information on to the
new thread so that it can initialize the GIL for itself.
Pseudo code ahead:
void do_in_thread(func_t *what) {
PyThreadState* state = PyThreadState_Get(); /// or new
PyInterpreterState_Current();
PyInterpreterState *interpreter = state->interp;
std::thread t([what, interpreter] {
auto s = PyGILState_EnsureEx(interpreter);
what();
PyGILState_Release(s); // could also release before what() because
TLS was updated and next PyGILState_Ensure() will work
});
}
Did I get that right?
Stephan
Am Mo., 28. Jan. 2019 um 09:27 Uhr schrieb Nick Coghlan :
>
> On Mon, 28 Jan 2019 at 00:32, Stephan Reiter wrote:
> >
> > Cool. Thanks, Nick!
> >
> > I did experiments based on this idea
> > (https://github.com/stephanreiter/cpython/commit/3bca91c26ac81e517b4aa22302be1741b3315622)
> > and haven't rejected it yet. :-)
>
> After talking to Graham about this, I unfortunately realised that the
> reason the callback approach is appearing to work for you is because
> your application is single-threaded, so you can readily map any
> invocation of the callback to the desired interpreter. Multi-threaded
> applications won't have that luxury - they need to be able to set the
> callback target on a per-thread basis.
>
> Graham actually described a plausible approach for doing that several
> years back: https://bugs.python.org/issue10915#msg126387
>
> We have much better subinterpreter testing support now, so if this is
> any area that you're interested in, one potential place to start would
> be to get Antoine's patch back to a point where it applies and
> compiles again.
>
> Cheers,
> Nick.
>
> --
> Nick Coghlan | ncogh...@gmail.com | Brisbane, Australia
___
Python-Dev mailing list
Python-Dev@python.org
https://mail.python.org/mailman/listinfo/python-dev
Unsubscribe:
https://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] Sub-interpreters: importing numpy causes hang
Cool. Thanks, Nick!
I did experiments based on this idea (
https://github.com/stephanreiter/cpython/commit/3bca91c26ac81e517b4aa22302be1741b3315622)
and haven't rejected it yet. :-)
Together with the other fix (
https://github.com/stephanreiter/cpython/commit/c1afa0c8cdfab862f409f1c7ff02b189f5191cbe),
numpy at least is happy in my Python-hosting app.
I will pursue the idea of swapping sys.modules in a single Interpreter now
because that wouldn't require patching Python and I might get the mileage
out of this approach I need.
Still interested in improving sub-interpreters, though. I just need to
balance short and long term solution. :-)
Stephan
Den søn. 27. jan. 2019, 15.17 skrev Nick Coghlan On Thu, 24 Jan 2019 at 05:45, Stephan Reiter
> wrote:
> > If we create a fresh OS thread and make it call PyGILState_Ensure, it
> > won't have a PyThreadState saved under autoTLSkey. That means it will
> > create one using the main interpreter. I, as the developer embedding
> > Python into my application and using multiple interpreters, have no
> > control here. Maybe I know that under current conditions a certain
> > other interpreter should be used.
> >
> > I'll try to provoke this situation and then introduce a callback from
> > Python into my application that will allow me to specify which
> > interpreter should be used, e.g. code as follows:
> >
> > PyInterpreter *pickAnInterpreter() {
> > return activePlugin ? activePlugin->interpreter : nullptr; //
> > nullptr maps to main interpreter
> > }
> >
> > PyGILState_SetNewThreadInterpreterSelectionCallback();
> >
> > Maybe rubbish. But I think a valuable experiment that will give me a
> > better understanding.
>
> That actually sounds like a pretty plausible approach to me, at least
> for cases where the embedding application maintains some other state
> that lets it know which interpreter a new thread should be associated
> with. The best aspect of it is that it would let the embedding
> application decide how to handle registration of previously unknown
> threads with the Python runtime *without* requiring that all existing
> extension modules switch to a new thread registration API first.
>
> I'll pass the concept along to Graham Dumpleton (author of the
> mod_wsgi module for Apache httpd) to see if an interface like this
> might be enough to resolve some of the major compatibility issues
> mod_wsgi currently encounters with subinterpreters.
>
> Cheers,
> Nick.
>
> --
> Nick Coghlan | ncogh...@gmail.com | Brisbane, Australia
>
___
Python-Dev mailing list
Python-Dev@python.org
https://mail.python.org/mailman/listinfo/python-dev
Unsubscribe:
https://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
Re: [Python-Dev] Sub-interpreters: importing numpy causes hang
You all do make me feel very welcome in this community! Thank you very much! :-)
And thank you for all the thought and time you put into your message,
Eric. I do appreciate in particular all the alternatives you
presented; you provide a good picture of my options.
Not ruling out any of them, I'll stick with (single process + multiple
subinterpreters + plugins can't keep state in Python + all my Python
calls are performed on the main thread) for the time being. That's
quite a limited environment, which I hope I can make work in the long
run. And I think the concept of subinterpreters is nice and I'd like
to spend some time on the challenge of improving the situation.
So, I updated my changes and have the following on top of 3.6.1 at the moment:
https://github.com/stephanreiter/cpython/commit/c1afa0c8cdfab862f409f1c7ff02b189f5191cbe
I did what Henry suggested and ran the Python test suite. On Windows,
with my changes I get as output:
357 tests OK.
2 tests failed:
test_re test_subprocess
46 tests skipped:
test_bz2 test_crypt test_curses test_dbm_gnu test_dbm_ndbm
test_devpoll test_epoll test_fcntl test_fork1 test_gdb test_grp
test_idle test_ioctl test_kqueue test_lzma test_nis test_openpty
test_ossaudiodev test_pipes test_poll test_posix test_pty test_pwd
test_readline test_resource test_smtpnet test_socketserver
test_spwd test_sqlite test_ssl test_syslog test_tcl
test_threadsignals test_timeout test_tix test_tk test_ttk_guionly
test_ttk_textonly test_turtle test_urllib2net test_urllibnet
test_wait3 test_wait4 test_winsound test_xmlrpc_net test_zipfile64
Total duration: 6 min 20 sec
Tests result: FAILURE
I dropped my changes and ran the test suite again using vanilla Python
and got the same result.
So, it seems that the change doesn't break anything that is tested,
but that probably doesn't mean a lot.
Tomorrow, I'll investigate the following situation if I find time:
If we create a fresh OS thread and make it call PyGILState_Ensure, it
won't have a PyThreadState saved under autoTLSkey. That means it will
create one using the main interpreter. I, as the developer embedding
Python into my application and using multiple interpreters, have no
control here. Maybe I know that under current conditions a certain
other interpreter should be used.
I'll try to provoke this situation and then introduce a callback from
Python into my application that will allow me to specify which
interpreter should be used, e.g. code as follows:
PyInterpreter *pickAnInterpreter() {
return activePlugin ? activePlugin->interpreter : nullptr; //
nullptr maps to main interpreter
}
PyGILState_SetNewThreadInterpreterSelectionCallback();
Maybe rubbish. But I think a valuable experiment that will give me a
better understanding.
Stephan
Am Mi., 23. Jan. 2019 um 18:11 Uhr schrieb Eric Snow
:
>
> Hi Stephan,
>
> On Tue, Jan 22, 2019 at 9:25 AM Stephan Reiter
> wrote:
> > I am new to the list and arriving with a concrete problem that I'd
> > like to fix myself.
>
> That is great! Statements like that are a good way to get folks
> interested in your success. :)
>
> > I am embedding Python (3.6) into my C++ application and I would like
> > to run Python scripts isolated from each other using sub-interpreters.
> > I am not using threads; everything is supposed to run in the
> > application's main thread.
>
> FYI, running multiple interpreters in the same (e.g. main) thread
> isn't as well thought out as running them in separate threads. There
> may be assumptions in the runtime that would cause crashes or
> inconsistency in the runtime, so be vigilant. Is there a reason not
> to run the subinterpreters in separate threads?
>
> Regarding isolation, keep in mind that there are some limitations. At
> an intrinsic level subinterpreters are never truly isolated since they
> run in the same process. This matters if you have concerns about
> security (which you should always consider) and stability (if a
> subinterpreter crashes then your whole process crashes). You can find
> that complete isolation via subprocess & multiprocessing.
>
> On top of intrinsic isolation, currently subinterpreters have gaps in
> isolation that need fixing. For instance, they share a lot of
> module-global state, as well as builtin types and singletons. So data
> can leak between subinterpreters unexpectedly.
>
> Finally, at the Python level subinterpreters don't have a good way to
> pass data around. (I'm working on that. [1]) Naturally at the C
> level you can keep pointers to objects and share data that way. Just
> keep in mind that doing so relies on the GIL (in an
> interpreter-per-thread scenario, which you're avoiding). In a world
> where subinterpreters don't share the GIL [2] (and you're running one
> interpreter per thread) you'll end up with refcounti
Re: [Python-Dev] Sub-interpreters: importing numpy causes hang
Hi!
Well, the plugins would be created by third-parties and I'd like them
to enable bunding of modules with their plugins.
I am afraid of modules with the same name, but being different, or
different versions of modules being used by different plugins. If
plugins share an interpreter, the module with a given name that is
imported first sticks around forever and for all plugins.
I am thinking about this design:
- Plugins don't maintain state in their Python world. They expose
functions, my application calls them.
- Everytime I call into them, they are presented with a clean global
namespace. After the call, the namespace (dict) is thrown away. That
releases any objects the plugin code has created.
- So, then I could also actively unload modules they loaded. But I do
know that this is problematic in particular for modules that use
native code.
I am interested in both a short-term and a long-term solution.
Actually, making subinterpreters work better is pretty sexy ...
because it's hard. :-)
Stephan
Am Mi., 23. Jan. 2019 um 11:30 Uhr schrieb Petr Viktorin :
>
> On 1/23/19 3:33 AM, Stephan Reiter wrote:
> > Thanks for the answers so far. I appreciate them!
> >
> > Nathaniel, I'd like to allow Python plugins in my application. A
> > plugin should be allowed to bring its own modules along (i.e.
> > plugin-specific subdir is in sys.path when the plugin is active) and
> > hence some isolation of them will be needed, so that they can use
> > different versions of a given module. That's my main motivation for
> > using subinterpreters.
> > I thought about running plugins out-of-processes - a separate process
> > for every plugin - and allow them to communicate with my application
> > via RPC. But that makes it more complex to implement the API my
> > application will offer and will slow down things due to the need to
> > copy data.
> > Maybe you have another idea for me? :)
>
> Try to make the plugins work together. Look into using pip/PyPI for your
> plugins. Try to make it so each package ("plugin") would have only one
> module/package, and dependencies would be other packages that can be
> installed individually and shared. And keep in mind you can set up your
> own package index, or distribute/install individual package files.
>
> If that's not possible, and you want things to work now, go with subprocess.
>
> If you want to help make subinterpreters work better, there are several
> people scratching at the problem from different angles. Most/all would
> welcome help, but don't expect any short-term benefits.
> (FWIW, my own effort is currently blocked on PEP 580, and I hope to move
> forward after a Council is elected.)
>
>
> > Henry, Antoine, thanks for your input; I'll check out the tests and
> > see what I can learn from issue 10915.
> >
> > Stephan
> >
> > Am Di., 22. Jan. 2019 um 22:39 Uhr schrieb Nathaniel Smith :
> >>
> >> There are currently numerous incompatibilities between numpy and
> >> subinterpreters, and no concrete plan for fixing them. The numpy team does
> >> not consider subinterpreters to be a supported configuration, and can't
> >> help you with any issues you run into. I know the concept of
> >> subinterpreters is really appealing, but unfortunately the CPython
> >> implementation is not really mature or widely supported... are you
> >> absolutely certain you need to use subinterpreters for your application?
> >>
> >> On Tue, Jan 22, 2019, 08:27 Stephan Reiter >>>
> >>> Hi all!
> >>>
> >>> I am new to the list and arriving with a concrete problem that I'd
> >>> like to fix myself.
> >>>
> >>> I am embedding Python (3.6) into my C++ application and I would like
> >>> to run Python scripts isolated from each other using sub-interpreters.
> >>> I am not using threads; everything is supposed to run in the
> >>> application's main thread.
> >>>
> >>> I noticed that if I create an interpreter, switch to it and execute
> >>> code that imports numpy (1.13), my application will hang.
> >>>
> >>>ntdll.dll!NtWaitForSingleObject() Unknown
> >>>KernelBase.dll!WaitForSingleObjectEx() Unknown
> >>>> python36.dll!_PyCOND_WAIT_MS(_PyCOND_T * cv=0x748a67a0,
> >>>> _RTL_CRITICAL_SECTION * cs=0x748a6778, unsigned long ms=5) Line
> >>>> 245 C
> >>>[Inline Frame] python36.dll!PyCOND_TIMEDWAIT(_PyCOND_T *) Line 275 C
> >>>python36.dll!take_gil(_ts * tstate=0x023251cbc260) Line 224 C
> >>>python36.dll!PyEval_Resto
Re: [Python-Dev] Sub-interpreters: importing numpy causes hang
Thanks for the answers so far. I appreciate them! Nathaniel, I'd like to allow Python plugins in my application. A plugin should be allowed to bring its own modules along (i.e. plugin-specific subdir is in sys.path when the plugin is active) and hence some isolation of them will be needed, so that they can use different versions of a given module. That's my main motivation for using subinterpreters. I thought about running plugins out-of-processes - a separate process for every plugin - and allow them to communicate with my application via RPC. But that makes it more complex to implement the API my application will offer and will slow down things due to the need to copy data. Maybe you have another idea for me? :) Henry, Antoine, thanks for your input; I'll check out the tests and see what I can learn from issue 10915. Stephan Am Di., 22. Jan. 2019 um 22:39 Uhr schrieb Nathaniel Smith : > > There are currently numerous incompatibilities between numpy and > subinterpreters, and no concrete plan for fixing them. The numpy team does > not consider subinterpreters to be a supported configuration, and can't help > you with any issues you run into. I know the concept of subinterpreters is > really appealing, but unfortunately the CPython implementation is not really > mature or widely supported... are you absolutely certain you need to use > subinterpreters for your application? > > On Tue, Jan 22, 2019, 08:27 Stephan Reiter > >> Hi all! >> >> I am new to the list and arriving with a concrete problem that I'd >> like to fix myself. >> >> I am embedding Python (3.6) into my C++ application and I would like >> to run Python scripts isolated from each other using sub-interpreters. >> I am not using threads; everything is supposed to run in the >> application's main thread. >> >> I noticed that if I create an interpreter, switch to it and execute >> code that imports numpy (1.13), my application will hang. >> >> ntdll.dll!NtWaitForSingleObject() Unknown >> KernelBase.dll!WaitForSingleObjectEx() Unknown >> > python36.dll!_PyCOND_WAIT_MS(_PyCOND_T * cv=0x748a67a0, >> > _RTL_CRITICAL_SECTION * cs=0x748a6778, unsigned long ms=5) Line >> > 245 C >> [Inline Frame] python36.dll!PyCOND_TIMEDWAIT(_PyCOND_T *) Line 275 C >> python36.dll!take_gil(_ts * tstate=0x023251cbc260) Line 224 C >> python36.dll!PyEval_RestoreThread(_ts * tstate=0x023251cbc260) Line >> 370 C >> python36.dll!PyGILState_Ensure() Line 855 C >> umath.cp36-win_amd64.pyd!7ff8c6306ab2() Unknown >> umath.cp36-win_amd64.pyd!7ff8c630723c() Unknown >> umath.cp36-win_amd64.pyd!7ff8c6303a1d() Unknown >> umath.cp36-win_amd64.pyd!7ff8c63077c0() Unknown >> umath.cp36-win_amd64.pyd!7ff8c62ff926() Unknown >> [Inline Frame] python36.dll!_PyObject_FastCallDict(_object *) Line 2316 C >> [Inline Frame] python36.dll!_PyObject_FastCallKeywords(_object *) Line >> 2480 C >> python36.dll!call_function(_object * * * >> pp_stack=0x0048be5f5e40, __int64 oparg, _object * kwnames) Line >> 4822 C >> >> Numpy's extension umath calls PyGILState_Ensure(), which in turn calls >> PyEval_RestoreThread on the (auto) threadstate of the main >> interpreter. And that's wrong. >> We are already holding the GIL with the threadstate of our current >> sub-interpreter, so there's no need to switch. >> >> I know that the GIL API is not fully compatible with sub-interpreters, >> as issues #10915 and #15751 illustrate. >> >> But since I need to support calls to PyGILState_Ensure - numpy is the >> best example -, I am trying to improve the situation here: >> https://github.com/stephanreiter/cpython/commit/d9d3451b038af2820f500843b6a88f57270e1597 >> >> That change may be naive, but it does the trick for my use case. If >> totally wrong, I don't mind pursuing another alley. >> >> Essentially, I'd like to ask for some guidance in how to tackle this >> problem while keeping the current GIL API unchanged (to avoid breaking >> modules). >> >> I am also wondering how I can test any changes I am proposing. Is >> there a test suite for interpreters, for example? >> >> Thank you very much, >> Stephan >> ___ >> Python-Dev mailing list >> Python-Dev@python.org >> https://mail.python.org/mailman/listinfo/python-dev >> Unsubscribe: >> https://mail.python.org/mailman/options/python-dev/njs%40pobox.com ___ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
[Python-Dev] Sub-interpreters: importing numpy causes hang
Hi all! I am new to the list and arriving with a concrete problem that I'd like to fix myself. I am embedding Python (3.6) into my C++ application and I would like to run Python scripts isolated from each other using sub-interpreters. I am not using threads; everything is supposed to run in the application's main thread. I noticed that if I create an interpreter, switch to it and execute code that imports numpy (1.13), my application will hang. ntdll.dll!NtWaitForSingleObject() Unknown KernelBase.dll!WaitForSingleObjectEx() Unknown > python36.dll!_PyCOND_WAIT_MS(_PyCOND_T * cv=0x748a67a0, > _RTL_CRITICAL_SECTION * cs=0x748a6778, unsigned long ms=5) Line 245 C [Inline Frame] python36.dll!PyCOND_TIMEDWAIT(_PyCOND_T *) Line 275 C python36.dll!take_gil(_ts * tstate=0x023251cbc260) Line 224 C python36.dll!PyEval_RestoreThread(_ts * tstate=0x023251cbc260) Line 370 C python36.dll!PyGILState_Ensure() Line 855 C umath.cp36-win_amd64.pyd!7ff8c6306ab2() Unknown umath.cp36-win_amd64.pyd!7ff8c630723c() Unknown umath.cp36-win_amd64.pyd!7ff8c6303a1d() Unknown umath.cp36-win_amd64.pyd!7ff8c63077c0() Unknown umath.cp36-win_amd64.pyd!7ff8c62ff926() Unknown [Inline Frame] python36.dll!_PyObject_FastCallDict(_object *) Line 2316 C [Inline Frame] python36.dll!_PyObject_FastCallKeywords(_object *) Line 2480 C python36.dll!call_function(_object * * * pp_stack=0x0048be5f5e40, __int64 oparg, _object * kwnames) Line 4822 C Numpy's extension umath calls PyGILState_Ensure(), which in turn calls PyEval_RestoreThread on the (auto) threadstate of the main interpreter. And that's wrong. We are already holding the GIL with the threadstate of our current sub-interpreter, so there's no need to switch. I know that the GIL API is not fully compatible with sub-interpreters, as issues #10915 and #15751 illustrate. But since I need to support calls to PyGILState_Ensure - numpy is the best example -, I am trying to improve the situation here: https://github.com/stephanreiter/cpython/commit/d9d3451b038af2820f500843b6a88f57270e1597 That change may be naive, but it does the trick for my use case. If totally wrong, I don't mind pursuing another alley. Essentially, I'd like to ask for some guidance in how to tackle this problem while keeping the current GIL API unchanged (to avoid breaking modules). I am also wondering how I can test any changes I am proposing. Is there a test suite for interpreters, for example? Thank you very much, Stephan ___ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/archive%40mail-archive.com
