gh-103533: Use pep669 APIs for cprofile #103534

gaogaotiantian · 2023-04-14T06:32:11Z

Issue: Use PEP 669 API for cprofile #103533

use the profiler tool id

gaogaotiantian · 2023-04-16T03:34:57Z

@ericsnowcurrently how can I declare a global constant table in the module?

ericsnowcurrently · 2023-04-17T15:03:52Z

@ericsnowcurrently how can I declare a global constant table in the module?

Is it a new table or is it an existing one you're trying to convert? Does it have objects in it? Is the data actually const?

gaogaotiantian · 2023-04-17T17:19:12Z

Is it a new table or is it an existing one you're trying to convert? Does it have objects in it? Is the data actually const?

It's a new, pure C, true const table.

static const CallbackTableEntry callback_table[] = {
    {PY_MONITORING_EVENT_PY_START, "_pystart_callback"},
    {PY_MONITORING_EVENT_PY_RESUME, "_pystart_callback"},
    {PY_MONITORING_EVENT_PY_RETURN, "_pyreturn_callback"},
    {PY_MONITORING_EVENT_PY_YIELD, "_pyreturn_callback"},
    {PY_MONITORING_EVENT_PY_UNWIND, "_pyreturn_callback"},
    {PY_MONITORING_EVENT_CALL, "_ccall_callback"},
    {PY_MONITORING_EVENT_C_RETURN, "_creturn_callback"},
    {PY_MONITORING_EVENT_C_RAISE, "_creturn_callback"},
    {0, NULL}
};

gaogaotiantian · 2023-04-20T01:26:04Z

The situation is almost identical to error_codes[] in Modules/_sqlite/module.c, so I did the same thing - add the variable to ignored.tsv.

gaogaotiantian · 2023-05-02T16:59:24Z

Hi @markshannon , do you think this is a good candidate for 3.12? We pushed out PEP 699 but none of the standard library actually uses it. The profiling tool change is simpler than the debugging tool, so maybe this could be an example/try out for implementing tools in PEP 669. Potentially we can have more feedbacks for the monitoring mechanism.

markshannon

I think we can skip creating builtin functions when profiling method descriptors.
Other than that, looks good.

Did you measure performance at all?

markshannon · 2023-05-04T12:56:58Z

Modules/_lsprof.c

+        Py_INCREF(callable);
+        return (PyObject*)((PyCFunctionObject *)callable);
+    }
+    if (Py_TYPE(callable) == &PyMethodDescr_Type) {


Is this necessary?
Doesn't the profiler extract the same data from the builtin function that it could from the method descriptor?

This piece is copied from the new setprofile I believe. The idea behind it is to make sure get_cfunc_from_callable only returns a PyCFunctionObject. If it's not, then ((PyCFunctionObject *)cfunc)->m_ml won't work.

What do you propose here? Simply return callable? We need to check that anyway because CALL event can be triggered before calling a Python function and we don't want to add profiler entry on that(it should be dealt with later). I did realize that Py_RETURN_NONE was incorrect - NULL should be returned.

All that happens to the builtin function objects is it gets passed down to normalizeUserObj() which then does some elaborate lookup to get the method descriptor back again. Both the method descriptor and builtin function contain a pointer to the same PyMethodDef struct.

So, leave the method descriptor alone here, and in normalizeUserObj() create the same string that as would be created for the builtin function.

The mo object on line 175 is the method descriptor.

There are three purposes get_cfunc_from_callable needs the serve:

To get the object(callable) to print later, which could be resolved in normalizeUserObj() like you said.

To get a unique key for the hash table(actually a binary tree). In this case, ((PyCFunctionObject *)cfunc)->m_ml was used, maybe the callable itself could be used as well? Not sure why m_ml was chosen at the first place, the builtin functions probably have dinstinct addresses anyway? Even for the builtin methods(like list.append), using (void*)callable directly might work?

To filter out unwanted data. This is why this part is necessary. Without trying to get the C function from the method, how could we know if this is a call to a builtin function that we would like to log? It could be a simple Python call(f()), which would trigger CALL event. It could be a method call, but not to a C function(self.some_method()). We need to filter these entry out before we even reach normalizeUserObj()(that's where we are about to log the function call).

The current implementation is trying to swap out the setprofile layer without touching the internal profiling system. It's true that the profiling system could be optimized, but it's also risky and probably need extra care. If we want to land this in 3.12, maybe we should avoid changing the profiling logic for now.

I agree with your point about risk.

I think it is worth cleaning up the internals. We couldn't do it before, as the conversion from method descriptor to builtin function occurred before cprofile got to see it. Maybe we can get it done for 3.12, maybe not.

So let's get this change in for 3.12, and we can streamline things later.

We can clean up the internals, but we still need to address point 3. We need to filter out entries that we don't want, and that probably requires resolving the descriptors.

All the information in the fake builtin function is also in the original method descriptor, so whatever the filter was doing should still work.

Do you suggest that we can get rid of Py_TYPE(callable)->tp_descr_get(callable, self_arg, (PyObject*)Py_TYPE(self_arg)) because we don't need the actual PyCFunctionObject in it? We still need to check against Py_TYPE(callable) == &PyMethodDescr_Type for the actual "builtin methods" right? To filter out Python defined methods?

gaogaotiantian · 2023-05-04T17:59:48Z

Did you measure performance at all?

The performance measurement was in the original issue #103533, it might not be the most obvious format...

gaogaotiantian · 2023-05-04T18:01:19Z

I guess we don't have time now to make C level API changes if we want to land this in 3.12, but it would be nice to have MISSING and DISABLE available for C code.

markshannon

One last question

markshannon · 2023-05-05T17:14:27Z

Lib/test/test_cprofile.py

@@ -25,7 +25,6 @@ def test_bad_counter_during_dealloc(self):
        with support.catch_unraisable_exception() as cm:
            obj = _lsprof.Profiler(lambda: int)
            obj.enable()
-            obj = _lsprof.Profiler(1)


Why has this line been removed?

Because what the test was partially testing won't work anymore. The line relies on the fact that the original obj was deallocated when replace - that won't happen because it's still being referenced by the monitors.

Having that line would also cause an infinite exception loop, probably due to the new mechanism.

Sorry not infinite, but very frequent warning. The issue is we did not disable the profiling - that's what I mentioned in the description - enabling the profiling with one object and disable it with another won't work.

* main: pythongh-99113: Add PyInterpreterConfig.own_gil (pythongh-104204) pythongh-104146: Remove unused var 'parser_body_declarations' from clinic.py (python#104214) pythongh-99113: Add Py_MOD_PER_INTERPRETER_GIL_SUPPORTED (pythongh-104205) pythongh-104108: Add the Py_mod_multiple_interpreters Module Def Slot (pythongh-104148) pythongh-99113: Share the GIL via PyInterpreterState.ceval.gil (pythongh-104203) pythonGH-100479: Add `pathlib.PurePath.with_segments()` (pythonGH-103975) pythongh-69152: Add _proxy_response_headers attribute to HTTPConnection (python#26152) pythongh-103533: Use PEP 669 APIs for cprofile (pythonGH-103534) pythonGH-96803: Add three C-API functions to make _PyInterpreterFrame less opaque for users of PEP 523. (pythonGH-96849)

Use pep669 APIs for cprofile

a59daf6

bedevere-bot mentioned this pull request Apr 14, 2023

Use PEP 669 API for cprofile #103533

Closed

bedevere-bot added the awaiting review label Apr 14, 2023

gaogaotiantian marked this pull request as ready for review April 14, 2023 06:32

📜🤖 Added by blurb_it.

82946fa

erlend-aasland requested a review from markshannon April 14, 2023 08:14

arhadthedev added stdlib Python modules in the Lib dir performance Performance or resource usage labels Apr 14, 2023

gaogaotiantian added 2 commits April 15, 2023 17:45

Clean up code. Report ValueError when another tool is trying to

3c88fd1

use the profiler tool id

Make the callback table const

2c9bed4

Make global check happy

20f4663

brandtbucher self-requested a review May 2, 2023 20:59

markshannon reviewed May 4, 2023

View reviewed changes

Fixed return None to NULL

b7b719d

markshannon reviewed May 5, 2023

View reviewed changes

markshannon merged commit b979741 into python:main May 5, 2023

bedevere-bot removed the awaiting review label May 5, 2023

gaogaotiantian deleted the pep669-cprofile branch May 5, 2023 17:56

jbower-fb pushed a commit to jbower-fb/cpython-jbowerfb that referenced this pull request May 8, 2023

pythongh-103533: Use PEP 669 APIs for cprofile (pythonGH-103534)

21fb990

This was referenced Oct 31, 2024

_lsprof.Profiler._creturn_callback() segfaults #126220

Closed

gh-126220: Adapt _lsprof to Argument Clinic #126233

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

gh-103533: Use pep669 APIs for cprofile #103534

gh-103533: Use pep669 APIs for cprofile #103534

gaogaotiantian commented Apr 14, 2023 •

edited by bedevere-bot

Loading

gaogaotiantian commented Apr 16, 2023

ericsnowcurrently commented Apr 17, 2023

gaogaotiantian commented Apr 17, 2023

gaogaotiantian commented Apr 20, 2023

gaogaotiantian commented May 2, 2023

markshannon left a comment

markshannon May 4, 2023 •

edited

Loading

gaogaotiantian May 4, 2023

markshannon May 5, 2023

markshannon May 5, 2023

gaogaotiantian May 5, 2023

markshannon May 5, 2023

gaogaotiantian May 5, 2023

markshannon May 5, 2023

gaogaotiantian May 5, 2023

gaogaotiantian commented May 4, 2023

gaogaotiantian commented May 4, 2023

markshannon left a comment

markshannon May 5, 2023

gaogaotiantian May 5, 2023

gaogaotiantian May 5, 2023

gh-103533: Use pep669 APIs for cprofile #103534

gh-103533: Use pep669 APIs for cprofile #103534

Conversation

gaogaotiantian commented Apr 14, 2023 • edited by bedevere-bot Loading

gaogaotiantian commented Apr 16, 2023

ericsnowcurrently commented Apr 17, 2023

gaogaotiantian commented Apr 17, 2023

gaogaotiantian commented Apr 20, 2023

gaogaotiantian commented May 2, 2023

markshannon left a comment

Choose a reason for hiding this comment

markshannon May 4, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gaogaotiantian commented May 4, 2023

gaogaotiantian commented May 4, 2023

markshannon left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gaogaotiantian commented Apr 14, 2023 •

edited by bedevere-bot

Loading

markshannon May 4, 2023 •

edited

Loading