Skip to content

gh-115999: Add free-threaded specialization for UNPACK_SEQUENCE #126600

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
Nov 22, 2024
Merged

gh-115999: Add free-threaded specialization for UNPACK_SEQUENCE #126600

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
5e0d875
Add FT specialization
Eclips4 Nov 8, 2024
fa8ff2f
Add test case
Eclips4 Nov 8, 2024
d49e9e6
Address Matt's review
Eclips4 Nov 8, 2024
ea1abf3
Refactor UNPACK_SEQUENCE_LIST
mpage Nov 15, 2024
7ad46ad
Merge branch 'main' into ft-specialize-unpack-sequence
Eclips4 Nov 21, 2024
66a2d67
Align PR with recent changes
Eclips4 Nov 21, 2024
16093d6
Update Python/specialize.c
Eclips4 Nov 21, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion Include/internal/pycore_opcode_metadata.h
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

2 changes: 1 addition & 1 deletion Include/internal/pycore_uop_metadata.h
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

31 changes: 31 additions & 0 deletions Lib/test/test_opcache.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1256,6 +1256,37 @@ def g():
self.assert_no_opcode(g, "CONTAINS_OP")


@cpython_only
@requires_specialization_ft
def test_unpack_sequence(self):
def f():
for _ in range(100):
a, b = 1, 2
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sort of unrelated to this PR, but I'm surprised constant unpacking like this isn't peepholed into LOAD_CONST, STORE_FAST, LOAD_CONST, STORE_FAST. Maybe because it's actually a longer instruction sequence than the original? It certainly does less work.

@iritkatriel, thoughts?

self.assertEqual(a, 1)
self.assertEqual(b, 2)

f()
self.assert_specialized(f, "UNPACK_SEQUENCE_TWO_TUPLE")
self.assert_no_opcode(f, "UNPACK_SEQUENCE")

def g():
for _ in range(100):
a, = 1,
self.assertEqual(a, 1)

g()
self.assert_specialized(g, "UNPACK_SEQUENCE_TUPLE")
self.assert_no_opcode(g, "UNPACK_SEQUENCE")

def x():
for _ in range(100):
a, b = [1, 2]
self.assertEqual(a, 1)
self.assertEqual(b, 2)

x()
self.assert_specialized(x, "UNPACK_SEQUENCE_LIST")
self.assert_no_opcode(x, "UNPACK_SEQUENCE")

if __name__ == "__main__":
unittest.main()
22 changes: 14 additions & 8 deletions Python/bytecodes.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1381,15 +1381,15 @@ dummy_func(
};

specializing op(_SPECIALIZE_UNPACK_SEQUENCE, (counter/1, seq -- seq)) {
#if ENABLE_SPECIALIZATION
#if ENABLE_SPECIALIZATION_FT
if (ADAPTIVE_COUNTER_TRIGGERS(counter)) {
next_instr = this_instr;
_Py_Specialize_UnpackSequence(seq, next_instr, oparg);
DISPATCH_SAME_OPARG();
}
OPCODE_DEFERRED_INC(UNPACK_SEQUENCE);
ADVANCE_ADAPTIVE_COUNTER(this_instr[1].counter);
#endif /* ENABLE_SPECIALIZATION */
#endif /* ENABLE_SPECIALIZATION_FT */
(void)seq;
(void)counter;
}
Expand Down Expand Up @@ -1429,12 +1429,18 @@ dummy_func(
inst(UNPACK_SEQUENCE_LIST, (unused/1, seq -- values[oparg])) {
PyObject *seq_o = PyStackRef_AsPyObjectBorrow(seq);
DEOPT_IF(!PyList_CheckExact(seq_o));
DEOPT_IF(PyList_GET_SIZE(seq_o) != oparg);
STAT_INC(UNPACK_SEQUENCE, hit);
PyObject **items = _PyList_ITEMS(seq_o);
for (int i = oparg; --i >= 0; ) {
*values++ = PyStackRef_FromPyObjectNew(items[i]);
int should_deopt = 0;
Py_BEGIN_CRITICAL_SECTION(seq_o);
should_deopt = PyList_GET_SIZE(seq_o) != oparg;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could this be rewritten as something like:

    if (PyList_GET_SIZE(seq_o) != oparg) {
         END_CRITICAL_SECTION
         DEOPT_IF(true);
    }

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We can't use the macros because they introduce and close a new scope.

I don't think there's precedence in CPython for using the critical section functions manually (everything uses the macros), but if it's important that we maintain the structure you're suggesting we could do something like

#ifdef Py_GIL_DISABLED
PyCriticalSection _py_cs;
PyCriticalSection_Begin(&_py_cs, seq_o);
#endif
if (PyList_GET_SIZE(seq_o) != oparg) {
    #ifdef Py_GIL_DISABLED
    PyCriticalSection_End(&_py_cs);
    #endif
    DEOPT_IF(true);
}
STAT_INC(UNPACK_SEQUENCE, hit);
PyObject **items = _PyList_ITEMS(seq_o);
for (int i = oparg; --i >= 0; ) {
    *values++ = PyStackRef_FromPyObjectNew(items[i]);
}
#ifdef Py_GIL_DISABLED
PyCriticalSection_End(&_py_cs);
#endif
DECREF_INPUTS();

Note that the preprocessor guards are not necessary for correctness; the critical section functions are a no-op in default builds.

if (!should_deopt) {
STAT_INC(UNPACK_SEQUENCE, hit);
PyObject **items = _PyList_ITEMS(seq_o);
for (int i = oparg; --i >= 0; ) {
*values++ = PyStackRef_FromPyObjectNew(items[i]);
}
}
Py_END_CRITICAL_SECTION();
DEOPT_IF(should_deopt);
DECREF_INPUTS();
}

Expand Down Expand Up @@ -2516,7 +2522,7 @@ dummy_func(
}
OPCODE_DEFERRED_INC(CONTAINS_OP);
ADVANCE_ADAPTIVE_COUNTER(this_instr[1].counter);
#endif /* ENABLE_SPECIALIZATION */
#endif /* ENABLE_SPECIALIZATION_FT */
}

macro(CONTAINS_OP) = _SPECIALIZE_CONTAINS_OP + _CONTAINS_OP;
Expand Down
22 changes: 16 additions & 6 deletions Python/executor_cases.c.h
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

26 changes: 18 additions & 8 deletions Python/generated_cases.c.h
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

35 changes: 12 additions & 23 deletions Python/specialize.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2464,7 +2464,6 @@ _Py_Specialize_CompareOp(_PyStackRef lhs_st, _PyStackRef rhs_st, _Py_CODEUNIT *i
cache->counter = adaptive_counter_cooldown();
}

#ifdef Py_STATS
static int
unpack_sequence_fail_kind(PyObject *seq)
{
Expand All @@ -2476,46 +2475,36 @@ unpack_sequence_fail_kind(PyObject *seq)
}
return SPEC_FAIL_OTHER;
}
#endif // Py_STATS

void
_Py_Specialize_UnpackSequence(_PyStackRef seq_st, _Py_CODEUNIT *instr, int oparg)
{
PyObject *seq = PyStackRef_AsPyObjectBorrow(seq_st);

assert(ENABLE_SPECIALIZATION);
assert(ENABLE_SPECIALIZATION_FT);
assert(_PyOpcode_Caches[UNPACK_SEQUENCE] ==
INLINE_CACHE_ENTRIES_UNPACK_SEQUENCE);
_PyUnpackSequenceCache *cache = (_PyUnpackSequenceCache *)(instr + 1);
if (PyTuple_CheckExact(seq)) {
if (PyTuple_GET_SIZE(seq) != oparg) {
SPECIALIZATION_FAIL(UNPACK_SEQUENCE, SPEC_FAIL_EXPECTED_ERROR);
goto failure;
unspecialize(instr, SPEC_FAIL_EXPECTED_ERROR);
return;
}
if (PyTuple_GET_SIZE(seq) == 2) {
instr->op.code = UNPACK_SEQUENCE_TWO_TUPLE;
goto success;
specialize(instr, UNPACK_SEQUENCE_TWO_TUPLE);
return;
}
instr->op.code = UNPACK_SEQUENCE_TUPLE;
goto success;
specialize(instr, UNPACK_SEQUENCE_TUPLE);
return;
}
if (PyList_CheckExact(seq)) {
if (PyList_GET_SIZE(seq) != oparg) {
SPECIALIZATION_FAIL(UNPACK_SEQUENCE, SPEC_FAIL_EXPECTED_ERROR);
goto failure;
unspecialize(instr, SPEC_FAIL_EXPECTED_ERROR);
return;
}
instr->op.code = UNPACK_SEQUENCE_LIST;
goto success;
specialize(instr, UNPACK_SEQUENCE_LIST);
return;
}
SPECIALIZATION_FAIL(UNPACK_SEQUENCE, unpack_sequence_fail_kind(seq));
failure:
STAT_INC(UNPACK_SEQUENCE, failure);
instr->op.code = UNPACK_SEQUENCE;
cache->counter = adaptive_counter_backoff(cache->counter);
return;
success:
STAT_INC(UNPACK_SEQUENCE, success);
cache->counter = adaptive_counter_cooldown();
unspecialize(instr, unpack_sequence_fail_kind(seq));
}

#ifdef Py_STATS
Expand Down
Loading