Skip to content

gh-142659: Optimize set_swap_bodies for intersection_update #148155

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

Open
Siyet wants to merge 2 commits into
base: main
Choose a base branch
from
+51 −54
Open

gh-142659: Optimize set_swap_bodies for intersection_update #148155

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
554d5c1
gh-142659: Optimize set_swap_bodies for intersection_update
Apr 6, 2026
37a95e6
Add news entry for gh-142659
Apr 6, 2026
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
3 changes: 3 additions & 0 deletions Misc/NEWS.d/next/Core_and_Builtins/2026-04-06-11-32-52.gh-issue-142659.44riUp.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,3 @@
Optimize :meth:`set.intersection_update` by replacing the general-purpose
``set_swap_bodies()`` with a specialized ``set_replace_body()`` that skips
unnecessary atomic operations and checks for the temporary set argument.
102 changes: 48 additions & 54 deletions Objects/setobject.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1475,74 +1475,68 @@ copy_small_table(setentry *dest, setentry *src)
}
#endif

/* set_swap_bodies() switches the contents of any two sets by moving their
internal data pointers and, if needed, copying the internal smalltables.
Semantically equivalent to:
/* set_replace_body() replaces the contents of dst with those of src,
moving dst's old contents into src for proper cleanup on Py_DECREF.

t=set(a); a.clear(); a.update(b); b.clear(); b.update(t); del t
The caller guarantees that src is a uniquely-referenced temporary set
Copy link
Copy Markdown
Contributor

@eendebakpt eendebakpt Apr 6, 2026

Choose a reason for hiding this comment

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

If the called is required to cleanup src immediately afterwards, maybe it is cleaner to let the set_replace_body steal the reference and do the decref.

that will be discarded immediately afterward. This allows us to skip
atomic operations and shared-marking on src's fields, and to skip the
frozenset hash swap (neither argument is ever a frozenset here).

The function always succeeds and it leaves both objects in a stable state.
Useful for operations that update in-place (by allowing an intermediate
result to be swapped into one of the original inputs).
*/

static void
set_swap_bodies(PySetObject *a, PySetObject *b)
set_replace_body(PySetObject *dst, PySetObject *src)
{
Py_ssize_t t;
setentry *u;
setentry tab[PySet_MINSIZE];
Py_hash_t h;

setentry *a_table = a->table;
setentry *b_table = b->table;
FT_ATOMIC_STORE_PTR_RELEASE(a->table, NULL);
FT_ATOMIC_STORE_PTR_RELEASE(b->table, NULL);

t = a->fill; a->fill = b->fill; b->fill = t;
t = a->used;
FT_ATOMIC_STORE_SSIZE_RELAXED(a->used, b->used);
FT_ATOMIC_STORE_SSIZE_RELAXED(b->used, t);
t = a->mask;
FT_ATOMIC_STORE_SSIZE_RELEASE(a->mask, b->mask);
FT_ATOMIC_STORE_SSIZE_RELEASE(b->mask, t);

u = a_table;
if (a_table == a->smalltable)
u = b->smalltable;
a_table = b_table;
if (b_table == b->smalltable)
a_table = a->smalltable;
b_table = u;

if (a_table == a->smalltable || b_table == b->smalltable) {
memcpy(tab, a->smalltable, sizeof(tab));

assert(!PyType_IsSubtype(Py_TYPE(dst), &PyFrozenSet_Type));
assert(!PyType_IsSubtype(Py_TYPE(src), &PyFrozenSet_Type));
assert(Py_REFCNT(src) == 1);
Copy link
Copy Markdown
Contributor

@eendebakpt eendebakpt Apr 6, 2026

Choose a reason for hiding this comment

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

Suggested change
assert(Py_REFCNT(src) == 1);
assert(_PyObject_IsUniquelyReferenced(src));

Copy link
Copy Markdown
Contributor

@colesbury colesbury Apr 6, 2026

Choose a reason for hiding this comment

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

This assertion is going to be prone to gc.get_objects()/gc.get_referrers() issues like in #148180.

I'm not sure if we should do anything about that.


setentry *dst_table = dst->table;
setentry *src_table = src->table;
FT_ATOMIC_STORE_PTR_RELEASE(dst->table, NULL);
src->table = NULL;

t = dst->fill; dst->fill = src->fill; src->fill = t;
t = dst->used;
FT_ATOMIC_STORE_SSIZE_RELAXED(dst->used, src->used);
src->used = t;
t = dst->mask;
FT_ATOMIC_STORE_SSIZE_RELEASE(dst->mask, src->mask);
src->mask = t;

u = dst_table;
if (dst_table == dst->smalltable)
u = src->smalltable;
dst_table = src_table;
if (src_table == src->smalltable)
dst_table = dst->smalltable;
src_table = u;

if (dst_table == dst->smalltable || src_table == src->smalltable) {
memcpy(tab, dst->smalltable, sizeof(tab));
#ifndef Py_GIL_DISABLED
memcpy(a->smalltable, b->smalltable, sizeof(tab));
memcpy(b->smalltable, tab, sizeof(tab));
memcpy(dst->smalltable, src->smalltable, sizeof(tab));
memcpy(src->smalltable, tab, sizeof(tab));
#else
copy_small_table(a->smalltable, b->smalltable);
copy_small_table(b->smalltable, tab);
copy_small_table(dst->smalltable, src->smalltable);
memcpy(src->smalltable, tab, sizeof(tab));
#endif
}

if (PyType_IsSubtype(Py_TYPE(a), &PyFrozenSet_Type) &&
PyType_IsSubtype(Py_TYPE(b), &PyFrozenSet_Type)) {
h = FT_ATOMIC_LOAD_SSIZE_RELAXED(a->hash);
FT_ATOMIC_STORE_SSIZE_RELAXED(a->hash, FT_ATOMIC_LOAD_SSIZE_RELAXED(b->hash));
FT_ATOMIC_STORE_SSIZE_RELAXED(b->hash, h);
} else {
FT_ATOMIC_STORE_SSIZE_RELAXED(a->hash, -1);
FT_ATOMIC_STORE_SSIZE_RELAXED(b->hash, -1);
}
if (!SET_IS_SHARED(b) && SET_IS_SHARED(a)) {
SET_MARK_SHARED(b);
}
if (!SET_IS_SHARED(a) && SET_IS_SHARED(b)) {
SET_MARK_SHARED(a);
FT_ATOMIC_STORE_SSIZE_RELAXED(dst->hash, -1);

if (SET_IS_SHARED(dst)) {
SET_MARK_SHARED(src);
}
FT_ATOMIC_STORE_PTR_RELEASE(a->table, a_table);
FT_ATOMIC_STORE_PTR_RELEASE(b->table, b_table);

FT_ATOMIC_STORE_PTR_RELEASE(dst->table, dst_table);
src->table = src_table;
}

/*[clinic input]
Expand Down Expand Up @@ -1797,7 +1791,7 @@ set_intersection_update(PySetObject *so, PyObject *other)
tmp = set_intersection(so, other);
if (tmp == NULL)
return NULL;
set_swap_bodies(so, (PySetObject *)tmp);
set_replace_body(so, (PySetObject *)tmp);
Py_DECREF(tmp);
Py_RETURN_NONE;
}
Expand All @@ -1821,7 +1815,7 @@ set_intersection_update_multi_impl(PySetObject *so, PyObject * const *others,
if (tmp == NULL)
return NULL;
Py_BEGIN_CRITICAL_SECTION(so);
set_swap_bodies(so, (PySetObject *)tmp);
set_replace_body(so, (PySetObject *)tmp);
Py_END_CRITICAL_SECTION();
Py_DECREF(tmp);
Py_RETURN_NONE;
Expand Down
Loading