gh-144650: support Vectorcall for _decimal.Decimal's

Misc/NEWS.d/next/Library/2026-09-26-05-18-21.gh-issue-144650.zqHLOW.rst

@@ -0,0 +1,2 @@
+Use AC vectorcall support for the :class:`decimal.Decimal`.  This offers
+~1.15x speedup on creation of small instances.

Modules/_decimal/_decimal.c

@@ -3229,7 +3229,7 @@ dec_apply(PyObject *v, PyObject *context)
 /* 'v' can have any type accepted by the Decimal constructor. Attempt
    an exact conversion. If the result does not meet the restrictions
    for an mpd_t, fail with InvalidOperation. */
-static PyObject *
+static inline PyObject *
 PyDecType_FromObjectExact(PyTypeObject *type, PyObject *v, PyObject *context)
 {
     decimal_state *state = get_module_state_from_ctx(context);
@@ -3313,6 +3313,7 @@ PyDec_FromObject(PyObject *v, PyObject *context)
 }
 
 /*[clinic input]
+@vectorcall
 @classmethod
 _decimal.Decimal.__new__ as dec_new
 
@@ -3329,7 +3330,7 @@ trap is active.
 
 static PyObject *
 dec_new_impl(PyTypeObject *type, PyObject *value, PyObject *context)
-/*[clinic end generated code: output=35f48a40c65625ba input=5f8a0892d3fcef80]*/
+/*[clinic end generated code: output=35f48a40c65625ba input=0e5ca99183562cd9]*/
 {
     decimal_state *state = get_module_state_by_def(type);
     CONTEXT_CHECK_VA(state, context);
@@ -6190,6 +6191,7 @@ static PyType_Slot dec_slots[] = {
     {Py_tp_methods, dec_methods},
     {Py_tp_getset, dec_getsets},
     {Py_tp_new, dec_new},
+    {Py_tp_vectorcall, dec_vectorcall},
 
     // Number protocol
     {Py_nb_add, nm_mpd_qadd},