pythongh-120642: Move private PyCode APIs to the internal C API (python#120643)

* Move _Py_CODEUNIT and related functions to pycore_code.h.
* Move _Py_BackoffCounter to pycore_backoff.h.
* Move Include/cpython/optimizer.h content to pycore_optimizer.h.
* Remove Include/cpython/optimizer.h.
* Remove PyUnstable_Replace_Executor().

Rename functions:

* PyUnstable_GetExecutor() => _Py_GetExecutor()
* PyUnstable_GetOptimizer() => _Py_GetOptimizer()
* PyUnstable_SetOptimizer() => _Py_SetTier2Optimizer()
* PyUnstable_Optimizer_NewCounter() => _PyOptimizer_NewCounter()
* PyUnstable_Optimizer_NewUOpOptimizer() => _PyOptimizer_NewUOpOptimizer()

Author: vstinner

Include/Python.h

@@ -132,6 +132,5 @@
 #include "fileutils.h"
 #include "cpython/pyfpe.h"
 #include "cpython/tracemalloc.h"
-#include "cpython/optimizer.h"
 
 #endif /* !Py_PYTHON_H */

Include/cpython/code.h

@@ -24,58 +24,6 @@ typedef struct _Py_GlobalMonitors {
     uint8_t tools[_PY_MONITORING_UNGROUPED_EVENTS];
 } _Py_GlobalMonitors;
 
-typedef struct {
-    union {
-        struct {
-            uint16_t backoff : 4;
-            uint16_t value : 12;
-        };
-        uint16_t as_counter;  // For printf("%#x", ...)
-    };
-} _Py_BackoffCounter;
-
-/* Each instruction in a code object is a fixed-width value,
- * currently 2 bytes: 1-byte opcode + 1-byte oparg.  The EXTENDED_ARG
- * opcode allows for larger values but the current limit is 3 uses
- * of EXTENDED_ARG (see Python/compile.c), for a maximum
- * 32-bit value.  This aligns with the note in Python/compile.c
- * (compiler_addop_i_line) indicating that the max oparg value is
- * 2**32 - 1, rather than INT_MAX.
- */
-
-typedef union {
-    uint16_t cache;
-    struct {
-        uint8_t code;
-        uint8_t arg;
-    } op;
-    _Py_BackoffCounter counter;  // First cache entry of specializable op
-} _Py_CODEUNIT;
-
-
-/* These macros only remain defined for compatibility. */
-#define _Py_OPCODE(word) ((word).op.code)
-#define _Py_OPARG(word) ((word).op.arg)
-
-static inline _Py_CODEUNIT
-_py_make_codeunit(uint8_t opcode, uint8_t oparg)
-{
-    // No designated initialisers because of C++ compat
-    _Py_CODEUNIT word;
-    word.op.code = opcode;
-    word.op.arg = oparg;
-    return word;
-}
-
-static inline void
-_py_set_opcode(_Py_CODEUNIT *word, uint8_t opcode)
-{
-    word->op.code = opcode;
-}
-
-#define _Py_MAKE_CODEUNIT(opcode, oparg) _py_make_codeunit((opcode), (oparg))
-#define _Py_SET_OPCODE(word, opcode) _py_set_opcode(&(word), (opcode))
-
 
 typedef struct {
     PyObject *_co_code;

Include/cpython/optimizer.h

@@ -13,6 +13,18 @@ extern "C" {
 #include <stdbool.h>
 #include <stdint.h>
 
+
+typedef struct {
+    union {
+        struct {
+            uint16_t backoff : 4;
+            uint16_t value : 12;
+        };
+        uint16_t as_counter;  // For printf("%#x", ...)
+    };
+} _Py_BackoffCounter;
+
+
 /* 16-bit countdown counters using exponential backoff.
 
    These are used by the adaptive specializer to count down until

Include/internal/pycore_backoff.h

@@ -9,6 +9,50 @@ extern "C" {
 #endif
 
 #include "pycore_lock.h"        // PyMutex
+#include "pycore_backoff.h"     // _Py_BackoffCounter
+
+
+/* Each instruction in a code object is a fixed-width value,
+ * currently 2 bytes: 1-byte opcode + 1-byte oparg.  The EXTENDED_ARG
+ * opcode allows for larger values but the current limit is 3 uses
+ * of EXTENDED_ARG (see Python/compile.c), for a maximum
+ * 32-bit value.  This aligns with the note in Python/compile.c
+ * (compiler_addop_i_line) indicating that the max oparg value is
+ * 2**32 - 1, rather than INT_MAX.
+ */
+
+typedef union {
+    uint16_t cache;
+    struct {
+        uint8_t code;
+        uint8_t arg;
+    } op;
+    _Py_BackoffCounter counter;  // First cache entry of specializable op
+} _Py_CODEUNIT;
+
+
+/* These macros only remain defined for compatibility. */
+#define _Py_OPCODE(word) ((word).op.code)
+#define _Py_OPARG(word) ((word).op.arg)
+
+static inline _Py_CODEUNIT
+_py_make_codeunit(uint8_t opcode, uint8_t oparg)
+{
+    // No designated initialisers because of C++ compat
+    _Py_CODEUNIT word;
+    word.op.code = opcode;
+    word.op.arg = oparg;
+    return word;
+}
+
+static inline void
+_py_set_opcode(_Py_CODEUNIT *word, uint8_t opcode)
+{
+    word->op.code = opcode;
+}
+
+#define _Py_MAKE_CODEUNIT(opcode, oparg) _py_make_codeunit((opcode), (oparg))
+#define _Py_SET_OPCODE(word, opcode) _py_set_opcode(&(word), (opcode))
 
 
 // We hide some of the newer PyCodeObject fields behind macros.

Include/internal/pycore_code.h

@@ -30,6 +30,7 @@ extern "C" {
 #include "pycore_list.h"          // struct _Py_list_state
 #include "pycore_mimalloc.h"      // struct _mimalloc_interp_state
 #include "pycore_object_state.h"  // struct _py_object_state
+#include "pycore_optimizer.h"     // _PyOptimizerObject
 #include "pycore_obmalloc.h"      // struct _obmalloc_state
 #include "pycore_qsbr.h"          // struct _qsbr_state
 #include "pycore_tstate.h"        // _PyThreadStateImpl