gh-139109: JIT _EXIT_TRACE to ENTER_EXECUTOR rather than _DEOPT

Include/internal/pycore_optimizer.h

@@ -362,7 +362,7 @@ PyAPI_FUNC(int) _PyDumpExecutors(FILE *out);
 extern void _Py_ClearExecutorDeletionList(PyInterpreterState *interp);
 #endif
 
-int _PyJit_translate_single_bytecode_to_trace(PyThreadState *tstate, _PyInterpreterFrame *frame, _Py_CODEUNIT *next_instr, bool stop_tracing);
+int _PyJit_translate_single_bytecode_to_trace(PyThreadState *tstate, _PyInterpreterFrame *frame, _Py_CODEUNIT *next_instr, int stop_tracing_opcode);
 
 int
 _PyJit_TryInitializeTracing(PyThreadState *tstate, _PyInterpreterFrame *frame,

Lib/test/test_capi/test_opt.py

@@ -40,6 +40,17 @@ def get_first_executor(func):
             pass
     return None
 
+def get_all_executors(func):
+    code = func.__code__
+    co_code = code.co_code
+    execs = []
+    for i in range(0, len(co_code), 2):
+        try:
+            execs.append(_opcode.get_executor(code, i))
+        except ValueError:
+            pass
+    return execs
+
 
 def iter_opnames(ex):
     for item in ex:
@@ -2629,6 +2640,18 @@ def gen():
             next(g)
         """ % _testinternalcapi.SPECIALIZATION_THRESHOLD))
 
+    def test_executor_side_exits_create_another_executor(self):
+        def f():
+            for x in range(TIER2_THRESHOLD + 3):
+                for y in range(TIER2_THRESHOLD + 3):
+                    z = x + y
+
+        f()
+        # Inner loop warms up first.
+        # Outer loop warms up later, linking to the inner one.
+        # Therefore, at least two executors.
+        self.assertGreaterEqual(len(get_all_executors(f)), 2)
+
 
 def global_identity(x):
     return x

Python/bytecodes.c

@@ -5644,7 +5644,7 @@ dummy_func(
             bool stop_tracing = (opcode == WITH_EXCEPT_START ||
                 opcode == RERAISE || opcode == CLEANUP_THROW ||
                 opcode == PUSH_EXC_INFO || opcode == INTERPRETER_EXIT);
-            int full = !_PyJit_translate_single_bytecode_to_trace(tstate, frame, next_instr, stop_tracing);
+            int full = !_PyJit_translate_single_bytecode_to_trace(tstate, frame, next_instr, stop_tracing ? _DEOPT : 0);
             if (full) {
                 LEAVE_TRACING();
                 int err = stop_tracing_and_jit(tstate, frame);
@@ -5684,7 +5684,7 @@ dummy_func(
 #if _Py_TIER2
             assert(IS_JIT_TRACING());
             int opcode = next_instr->op.code;
-            _PyJit_translate_single_bytecode_to_trace(tstate, frame, NULL, true);
+            _PyJit_translate_single_bytecode_to_trace(tstate, frame, NULL, _EXIT_TRACE);
             LEAVE_TRACING();
             int err = stop_tracing_and_jit(tstate, frame);
             ERROR_IF(err < 0);

Python/optimizer.c

@@ -574,7 +574,7 @@ _PyJit_translate_single_bytecode_to_trace(
     PyThreadState *tstate,
     _PyInterpreterFrame *frame,
     _Py_CODEUNIT *next_instr,
-    bool stop_tracing)
+    int stop_tracing_opocde)
 {
 
 #ifdef Py_DEBUG
@@ -637,8 +637,8 @@ _PyJit_translate_single_bytecode_to_trace(
         goto full;
     }
 
-    if (stop_tracing) {
-        ADD_TO_TRACE(_DEOPT, 0, 0, target);
+    if (stop_tracing_opocde != 0) {
+        ADD_TO_TRACE(stop_tracing_opocde, 0, 0, target);
         goto done;
     }