gh-134584: Eliminate redundant refcounting from _STORE_ATTR_INSTANCE_VALUE (GH-142759)

Signed-off-by: Manjusaka <me@manjusaka.me>

Python/bytecodes.c

@@ -2575,7 +2575,7 @@ dummy_func(
             }
         }
 
-        op(_STORE_ATTR_INSTANCE_VALUE, (offset/1, value, owner --)) {
+        op(_STORE_ATTR_INSTANCE_VALUE, (offset/1, value, owner -- o)) {
             PyObject *owner_o = PyStackRef_AsPyObjectBorrow(owner);
 
             STAT_INC(STORE_ATTR, hit);
@@ -2589,7 +2589,8 @@ dummy_func(
                 _PyDictValues_AddToInsertionOrder(values, index);
             }
             UNLOCK_OBJECT(owner_o);
-            PyStackRef_CLOSE(owner);
+            o = owner;
+            DEAD(owner);
             Py_XDECREF(old_value);
         }
 
@@ -2597,7 +2598,8 @@ dummy_func(
             unused/1 +
             _GUARD_TYPE_VERSION_AND_LOCK +
             _GUARD_DORV_NO_DICT +
-            _STORE_ATTR_INSTANCE_VALUE;
+            _STORE_ATTR_INSTANCE_VALUE +
+            POP_TOP;
 
         op(_STORE_ATTR_WITH_HINT, (hint/1, value, owner --)) {
             PyObject *owner_o = PyStackRef_AsPyObjectBorrow(owner);

Python/executor_cases.c.h

@@ -8728,11 +8728,12 @@
             break;
         }
 
-        case _STORE_ATTR_INSTANCE_VALUE_r20: {
+        case _STORE_ATTR_INSTANCE_VALUE_r21: {
             CHECK_CURRENT_CACHED_VALUES(2);
             assert(WITHIN_STACK_BOUNDS_IGNORING_CACHE());
             _PyStackRef owner;
             _PyStackRef value;
+            _PyStackRef o;
             _PyStackRef _stack_item_0 = _tos_cache0;
             _PyStackRef _stack_item_1 = _tos_cache1;
             owner = _stack_item_1;
@@ -8750,14 +8751,19 @@
                 _PyDictValues_AddToInsertionOrder(values, index);
             }
             UNLOCK_OBJECT(owner_o);
+            o = owner;
+            stack_pointer[0] = o;
+            stack_pointer += 1;
+            ASSERT_WITHIN_STACK_BOUNDS(__FILE__, __LINE__);
             _PyFrame_SetStackPointer(frame, stack_pointer);
-            PyStackRef_CLOSE(owner);
             Py_XDECREF(old_value);
             stack_pointer = _PyFrame_GetStackPointer(frame);
-            _tos_cache0 = PyStackRef_ZERO_BITS;
+            _tos_cache0 = o;
             _tos_cache1 = PyStackRef_ZERO_BITS;
             _tos_cache2 = PyStackRef_ZERO_BITS;
-            SET_CURRENT_CACHED_VALUES(0);
+            SET_CURRENT_CACHED_VALUES(1);
+            stack_pointer += -1;
+            ASSERT_WITHIN_STACK_BOUNDS(__FILE__, __LINE__);
             assert(WITHIN_STACK_BOUNDS_IGNORING_CACHE());
             break;
         }

Python/generated_cases.c.h

@@ -10535,6 +10535,7 @@
             static_assert(INLINE_CACHE_ENTRIES_STORE_ATTR == 4, "incorrect cache size");
             _PyStackRef owner;
             _PyStackRef value;
+            _PyStackRef o;
             /* Skip 1 cache entry */
             // _GUARD_TYPE_VERSION_AND_LOCK
             {
@@ -10588,13 +10589,23 @@
                     _PyDictValues_AddToInsertionOrder(values, index);
                 }
                 UNLOCK_OBJECT(owner_o);
-                stack_pointer += -2;
+                o = owner;
+                stack_pointer[-2] = o;
+                stack_pointer += -1;
                 ASSERT_WITHIN_STACK_BOUNDS(__FILE__, __LINE__);
                 _PyFrame_SetStackPointer(frame, stack_pointer);
-                PyStackRef_CLOSE(owner);
                 Py_XDECREF(old_value);
                 stack_pointer = _PyFrame_GetStackPointer(frame);
             }
+            // _POP_TOP
+            {
+                value = o;
+                stack_pointer += -1;
+                ASSERT_WITHIN_STACK_BOUNDS(__FILE__, __LINE__);
+                _PyFrame_SetStackPointer(frame, stack_pointer);
+                PyStackRef_XCLOSE(value);
+                stack_pointer = _PyFrame_GetStackPointer(frame);
+            }
             DISPATCH();
         }
 

Python/optimizer_bytecodes.c

@@ -99,6 +99,11 @@ dummy_func(void) {
         GETLOCAL(oparg) = temp;
     }
 
+    op(_STORE_ATTR_INSTANCE_VALUE, (offset/1, value, owner -- o)) {
+        (void)value;
+        o = owner;
+    }
+
     op(_STORE_FAST, (value --)) {
         GETLOCAL(oparg) = value;
     }

Python/optimizer_cases.c.h

@@ -1885,8 +1885,17 @@
         }
 
         case _STORE_ATTR_INSTANCE_VALUE: {
-            CHECK_STACK_BOUNDS(-2);
-            stack_pointer += -2;
+            JitOptRef owner;
+            JitOptRef value;
+            JitOptRef o;
+            owner = stack_pointer[-1];
+            value = stack_pointer[-2];
+            uint16_t offset = (uint16_t)this_instr->operand0;
+            (void)value;
+            o = owner;
+            CHECK_STACK_BOUNDS(-1);
+            stack_pointer[-2] = o;
+            stack_pointer += -1;
             ASSERT_WITHIN_STACK_BOUNDS(__FILE__, __LINE__);
             break;
         }