diff --git a/Include/floatobject.h b/Include/floatobject.h
index 60ede40..6c11036 100644
--- a/Include/floatobject.h
+++ b/Include/floatobject.h
@@ -21,6 +21,12 @@
 #define PyFloat_Check(op) PyObject_TypeCheck(op, &PyFloat_Type)
 #define PyFloat_CheckExact(op) (Py_TYPE(op) == &PyFloat_Type)
 
+/* The str() precision PyFloat_STR_PRECISION is chosen so that in most cases,
+   the rounding noise created by various operations is suppressed, while
+   giving plenty of precision for practical use. */
+
+#define PyFloat_STR_PRECISION 12
+
 #ifdef Py_NAN
 #define Py_RETURN_NAN return PyFloat_FromDouble(Py_NAN)
 #endif