gh-100239: Specialize more binary operations using BINARY_OP_EXTEND (GH-128956)

Author: eendebakpt

Include/internal/pycore_bytesobject.h

@@ -14,6 +14,11 @@ extern PyObject* _PyBytes_FormatEx(
     PyObject *args,
     int use_bytearray);
 
+/* Concatenate two bytes objects. Used as the sq_concat slot and by the
+ * specializing interpreter. Unlike PyBytes_Concat(), this returns a new
+ * reference rather than modifying its first argument in place. */
+extern PyObject* _PyBytes_Concat(PyObject *a, PyObject *b);
+
 extern PyObject* _PyBytes_FromHex(
     PyObject *string,
     int use_bytearray);

Include/internal/pycore_code.h

@@ -503,6 +503,11 @@ typedef struct {
        aliased to either operand). Used by the tier 2 optimizer to enable
        inplace follow-up ops. */
     int result_unique;
+    /* Expected types of the left and right operands. Used by the tier 2
+       optimizer to eliminate _GUARD_BINARY_OP_EXTEND when the operand
+       types are already known. NULL means unknown/don't eliminate. */
+    PyTypeObject *lhs_type;
+    PyTypeObject *rhs_type;
 } _PyBinaryOpSpecializationDescr;
 
 /* Comparison bit masks. */

Include/internal/pycore_dict.h

@@ -87,6 +87,11 @@ typedef struct {
 extern PyDictKeysObject *_PyDict_NewKeysForClass(PyHeapTypeObject *);
 extern PyObject *_PyDict_FromKeys(PyObject *, PyObject *, PyObject *);
 
+/* Implementations of the `|` and `|=` operators for dict, used by the
+ * specializing interpreter. */
+extern PyObject *_PyDict_Or(PyObject *self, PyObject *other);
+extern PyObject *_PyDict_IOr(PyObject *self, PyObject *other);
+
 /* Gets a version number unique to the current state of the keys of dict, if possible.
  * Returns the version number, or zero if it was not possible to get a version number. */
 extern uint32_t _PyDictKeys_GetVersionForCurrentState(

Include/internal/pycore_list.h

@@ -15,6 +15,7 @@ extern "C" {
 PyAPI_FUNC(PyObject*) _PyList_Extend(PyListObject *, PyObject *);
 PyAPI_FUNC(PyObject) *_PyList_SliceSubscript(PyObject*, PyObject*);
 PyAPI_FUNC(PyObject *) _PyList_BinarySlice(PyObject *, PyObject *, PyObject *);
+PyAPI_FUNC(PyObject *) _PyList_Concat(PyObject *, PyObject *);
 extern void _PyList_DebugMallocStats(FILE *out);
 // _PyList_GetItemRef should be used only when the object is known as a list
 // because it doesn't raise TypeError when the object is not a list, whereas PyList_GetItemRef does.

Include/internal/pycore_tuple.h

@@ -28,6 +28,7 @@ PyAPI_FUNC(void) _PyStolenTuple_Free(PyObject *self);
 PyAPI_FUNC(PyObject *)_PyTuple_FromStackRefStealOnSuccess(const union _PyStackRef *, Py_ssize_t);
 PyAPI_FUNC(PyObject *)_PyTuple_FromArraySteal(PyObject *const *, Py_ssize_t);
 PyAPI_FUNC(PyObject *) _PyTuple_BinarySlice(PyObject *, PyObject *, PyObject *);
+PyAPI_FUNC(PyObject *) _PyTuple_Concat(PyObject *, PyObject *);
 
 PyAPI_FUNC(PyObject *) _PyTuple_FromPair(PyObject *, PyObject *);
 PyAPI_FUNC(PyObject *) _PyTuple_FromPairSteal(PyObject *, PyObject *);