Minor optimization for set object, swap bodies #142659

Activity

1. nascheme

   added

   type-featureA feature request or enhancementA feature request or enhancement

   on Dec 13, 2025
2. StanFromIreland

   added

   performancePerformance or resource usagePerformance or resource usage

   interpreter-core(Objects, Python, Grammar, and Parser dirs)(Objects, Python, Grammar, and Parser dirs)

   on Dec 13, 2025
3. brijkapadia commented on Mar 9, 2026

   

   brijkapadia

   on Mar 9, 2026

   Contributor

   More actions

   For reference, the two functions that call set_swap_bodies are set_intersection_update and set_intersection_update_multi_impl:

   set_intersection_update

   cpython/Objects/setobject.c

   Lines 1792 to 1803 in 63eaaf9

   	static PyObject *
   	set_intersection_update(PySetObject *so, PyObject *other)
   	{
   	PyObject *tmp;
   	tmp = set_intersection(so, other);
   	if (tmp == NULL)
   	return NULL;
   	set_swap_bodies(so, (PySetObject *)tmp);
   	Py_DECREF(tmp);
   	Py_RETURN_NONE;
   	}

   set_intersection_update_multi_impl

   cpython/Objects/setobject.c

   Lines 1813 to 1828 in 63eaaf9

   	static PyObject *
   	set_intersection_update_multi_impl(PySetObject *so, PyObject * const *others,
   	Py_ssize_t others_length)
   	/*[clinic end generated code: output=d768b5584675b48d input=782e422fc370e4fc]*/
   	{
   	PyObject *tmp;
   	tmp = set_intersection_multi_impl(so, others, others_length);
   	if (tmp == NULL)
   	return NULL;
   	Py_BEGIN_CRITICAL_SECTION(so);
   	set_swap_bodies(so, (PySetObject *)tmp);
   	Py_END_CRITICAL_SECTION();
   	Py_DECREF(tmp);
   	Py_RETURN_NONE;
   	}

   I spent time trying to implement this optimization and found no easy fix. I realized that in order to optimize set_swap_bodies, you would basically need to manually free memory. For example, you would need to individually free each key in the set a in order for the set object to still exit in memory but not any of its keys (which we want to replace with the keys of set b).

   However there are two small optimizations you could do:

   1. As @eendebakpt mentioned, you can add an assert and get rid of if statements (especially the ones that are never true)
   2. Since tmp is uniquely referenced in set_intersection_update and set_intersection_update_multi, do you need atomic operations on b in set_swap_bodies? This could possible speed up the free-threaded build.

   I am not sure if these changes are worth a PR though.

   Happy to dig deeper if you think it is worth it.
4. Siyet

   added 6 commits that reference this issue on Apr 6, 2026

   pythongh-142659: Optimize set_swap_bodies for intersection_update

   ...

   26ccdf4

   Add news entry for pythongh-142659

   d3f1a20

   pythongh-142659: Optimize set_swap_bodies for intersection_update

   ...

   554d5c1

   Add news entry for pythongh-142659

   37a95e6

   pythongh-142659: Move GIL/free-threading branch into copy_small_table

   ...

   100c491

   pythongh-142659: Use PySet_Check in set_replace_body assertions

   0c0a00a