gh-158816: Skip redundant hash table probe when deleting from split dicts

delitem_common() unconditionally called lookdict_index() to locate the index slot in dk_indices that points at the entry being removed.

That slot is only used by the combined branch, to overwrite it with DKIX_DUMMY.  In split tables the whole probe sequence was redundant work.

Move the call into the combined branch and drop the now-pointless assert() for split tables.

Microbenchmark (free-threaded build, -Og, no PGO; deleting keys from materialized instance dicts; medians over 12 alternating runs):

    del d[k]           71.2 -> 69.2 ns/op   (-2.8%)
    d.pop(k)           98.9 -> 96.1 ns/op   (-2.8%)
    colliding hashes   78.0 -> 74.4 ns/op   (-4.6%)

clang still does not sink the call into the branch at -O3 -DNDEBUG, so this is not specific to the local
-Og build.

test_dict, test_dictviews and test_dictcomps pass.

Author: hetaozdh

Misc/NEWS.d/next/Core_and_Builtins/2026-10-04-20-04-55.gh-issue-158816.9ulhWb.rst

@@ -0,0 +1,3 @@
+Speed up deleting items from split dictionaries (for example the
+materialized ``__dict__`` of an instance with inline values) by skipping a
+redundant hash table probe.

Objects/dictobject.c

@@ -2938,9 +2938,6 @@ delitem_common(PyDictObject *mp, Py_hash_t hash, Py_ssize_t ix,
 
     PyObject *old_key;
 
-    Py_ssize_t hashpos = lookdict_index(mp->ma_keys, hash, ix);
-    assert(hashpos >= 0);
-
     STORE_USED(mp, mp->ma_used - 1);
     if (_PyDict_HasSplitTable(mp)) {
         assert(old_value == mp->ma_values->values[ix]);
@@ -2951,6 +2948,9 @@ delitem_common(PyDictObject *mp, Py_hash_t hash, Py_ssize_t ix,
         ASSERT_CONSISTENT(mp);
     }
     else {
+        Py_ssize_t hashpos = lookdict_index(mp->ma_keys, hash, ix);
+        assert(hashpos >= 0);
+
         FT_ATOMIC_STORE_UINT32_RELAXED(mp->ma_keys->dk_version, 0);
         dictkeys_set_index(mp->ma_keys, hashpos, DKIX_DUMMY);
         if (DK_IS_UNICODE(mp->ma_keys)) {