cmaloney · Oct 6, 2026 · Oct 15, 2026 · eendebakpt · Oct 15, 2026 · cmaloney
diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2026-10-05-13-19-49.gh-issue-87613.JLTGBy.rst b/Misc/NEWS.d/next/Core_and_Builtins/2026-10-05-13-19-49.gh-issue-87613.JLTGBy.rst
@@ -0,0 +1,2 @@
+Speed up :class:`bytes` and :class:`bytearray` construction by adding
+:ref:`vectorcall` support.
diff --git a/Objects/bytearrayobject.c b/Objects/bytearrayobject.c
@@ -977,6 +977,7 @@ bytearray_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
 }
 
 /*[clinic input]
+@vectorcall
 bytearray.__init__
 
     source as arg: object = NULL
@@ -988,7 +989,7 @@ bytearray.__init__
 static int
 bytearray___init___impl(PyByteArrayObject *self, PyObject *arg,
                         const char *encoding, const char *errors)
-/*[clinic end generated code: output=4ce1304649c2f8b3 input=1141a7122eefd7b9]*/
+/*[clinic end generated code: output=4ce1304649c2f8b3 input=24ddb84055f432dd]*/
 {
     Py_ssize_t count;
     PyObject *it;
@@ -3027,6 +3028,7 @@ PyTypeObject PyByteArray_Type = {
     PyType_GenericAlloc,                /* tp_alloc */
     bytearray_new,                      /* tp_new */
     PyObject_Free,                      /* tp_free */
+    .tp_vectorcall = bytearray_vectorcall,
     .tp_version_tag = _Py_TYPE_VERSION_BYTEARRAY,
 };
 

diff --git a/Objects/bytesobject.c b/Objects/bytesobject.c
@@ -2918,6 +2918,7 @@ static PyObject *
 bytes_subtype_new(PyTypeObject *, PyObject *);
 
 /*[clinic input]
+@vectorcall
 @classmethod
 bytes.__new__ as bytes_new
 
@@ -2930,7 +2931,7 @@ bytes.__new__ as bytes_new
 static PyObject *
 bytes_new_impl(PyTypeObject *type, PyObject *x, const char *encoding,
                const char *errors)
-/*[clinic end generated code: output=1e0c471be311a425 input=f0a966d19b7262b4]*/
+/*[clinic end generated code: output=1e0c471be311a425 input=b0248d22e221a095]*/
 {
     PyObject *bytes;
     PyObject *func;
@@ -3316,6 +3317,7 @@ PyTypeObject PyBytes_Type = {
     bytes_type_alloc,                           /* tp_alloc */
     bytes_new,                                  /* tp_new */
     PyObject_Free,                              /* tp_free */
+    .tp_vectorcall = bytes_vectorcall,
     .tp_version_tag = _Py_TYPE_VERSION_BYTES,
     ._tp_iteritem = bytes_iteritem,
 };

diff --git a/Objects/clinic/bytearrayobject.c.h b/Objects/clinic/bytearrayobject.c.h
diff --git a/Objects/clinic/bytesobject.c.h b/Objects/clinic/bytesobject.c.h
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1,2 @@
		Speed up :class:`bytes` and :class:`bytearray` construction by adding
eendebakpt Oct 15, 2026 Copy link Copy Markdown Contributor Choose a reason for hiding this comment The reason will be displayed to describe this comment to others. Learn more. The vectorcall changes look good. But the optimizations need a second look. For example: `>>> import bisect >>> lst = [] >>> bisect.insort(lst, bytearray(b'abc')[1:], key=bytes) >>> lst [bytearray(b'')] # main: [bytearray(b'bc')]` For that reason maybe split the PR? cmaloney reacted with eyes emoji cmaloney Oct 15, 2026 Copy link Copy Markdown Contributor Author Choose a reason for hiding this comment The reason will be displayed to describe this comment to others. Learn more. Splitting the PR It looks like this gets broken because a lot of the stack uses "borrowed" references which make it look like it's uniquely referenced when it is still in use in multiple paths.... Still want to do the no-copy optimization and the similar one with `iconcat` but clearly still need some different tools.
		:ref:`vectorcall` support.