WIP

python · corona10 · Jun 22, 2025 · Jun 22, 2025 · Jun 22, 2025 · Jun 22, 2025
commit 46b423fcd354bd229d6e430860946e1fe166bb18
diff --git a/Include/internal/pycore_opcode_metadata.h b/Include/internal/pycore_opcode_metadata.h
diff --git a/Lib/test/test_capi/test_opt.py b/Lib/test/test_capi/test_opt.py
@@ -5,6 +5,8 @@
 import unittest
 import gc
 import os
+import random
+import string
 
 import _opcode
 
@@ -2362,6 +2364,23 @@ def testfunc(n):
         self.assertNotIn("_GUARD_TOS_INT", uops)
         self.assertNotIn("_GUARD_NOS_INT", uops)
 
+    def test_store_fast_pop_top_specialize_unicode(self):
+        def random_str(n):
+            return ''.join(random.choice(string.ascii_uppercase + string.digits) for _ in range(n))
+        def testfunc(n):
+            y = random_str(32)
+            for _ in range(n):
+                x = y + y # _POP_TOP
+                x = None # _POP_TOP_NOP
+
+        testfunc(TIER2_THRESHOLD)
+
+        ex = get_first_executor(testfunc)
+        self.assertIsNotNone(ex)
+        uops = get_opnames(ex)
+
+        self.assertIn("_POP_TOP_NOP", uops)
+
 def test_float_op_refcount_elimination(self): 
 def test_float_op_refcount_elimination(self): 
     def test_attr_promotion_failure(self):
         # We're not testing for any specific uops here, just
         # testing it doesn't crash.

diff --git a/Lib/test/test_sys.py b/Lib/test/test_sys.py
@@ -1671,7 +1671,7 @@ def func():
             INTERPRETER_FRAME = '9PihcP'
         else:
             INTERPRETER_FRAME = '9PhcP'
-        check(x, size('3PiccPPP' + INTERPRETER_FRAME + 'P'))
+        check(x, size('3PiccPPP' + INTERPRETER_FRAME + 'PP'))
         # function
         def func(): pass
         check(func, size('16Pi'))
@@ -1688,7 +1688,7 @@ def bar(cls):
             check(bar, size('PP'))
         # generator
         def get_gen(): yield 1
-        check(get_gen(), size('6P4c' + INTERPRETER_FRAME + 'P'))
+        check(get_gen(), size('6P4c' + INTERPRETER_FRAME + 'PP'))
         # iterator
         check(iter('abc'), size('lP'))
         # callable-iterator

@@ -773,23 +773,23 @@ dummy_func(
         macro(BINARY_OP_SUBTRACT_FLOAT) =
             _GUARD_TOS_FLOAT + _GUARD_NOS_FLOAT + unused/5 + _BINARY_OP_SUBTRACT_FLOAT;
 
-        pure op(_BINARY_OP_ADD_UNICODE, (left, right -- res)) {
+        pure op(_BINARY_OP_ADD_UNICODE, (left, right -- res, l, r)) {
             PyObject *left_o = PyStackRef_AsPyObjectBorrow(left);
             PyObject *right_o = PyStackRef_AsPyObjectBorrow(right);
             assert(PyUnicode_CheckExact(left_o));
             assert(PyUnicode_CheckExact(right_o));
 
             STAT_INC(BINARY_OP, hit);
             PyObject *res_o = PyUnicode_Concat(left_o, right_o);
-            PyStackRef_CLOSE_SPECIALIZED(right, _PyUnicode_ExactDealloc);
-            PyStackRef_CLOSE_SPECIALIZED(left, _PyUnicode_ExactDealloc);
             INPUTS_DEAD();
             ERROR_IF(res_o == NULL);
+            l = left;
+            r = right;
             res = PyStackRef_FromPyObjectSteal(res_o);
         }
 
         macro(BINARY_OP_ADD_UNICODE) =
-            _GUARD_TOS_UNICODE + _GUARD_NOS_UNICODE + unused/5 + _BINARY_OP_ADD_UNICODE;
+            _GUARD_TOS_UNICODE + _GUARD_NOS_UNICODE + unused/5 + _BINARY_OP_ADD_UNICODE + _POP_TOP_UNICODE + _POP_TOP_UNICODE;
 
         // This is a subtle one. It's a super-instruction for
         // BINARY_OP_ADD_UNICODE followed by STORE_FAST

@@ -1431,6 +1431,10 @@ optimize_and_assemble_code_unit(struct compiler_unit *u, PyObject *const_cache,
                                                  &optimized_instrs) < 0) {
         goto error;
     }
+    /* Reserve an extra word on the stack to ensure there is space for uops to
+       pass at least one item on the stack to a subsequent uop.
+    */
+    stackdepth++;
 
     /** Assembly **/
     co = _PyAssemble_MakeCodeObject(&u->u_metadata, const_cache, consts,

diff --git a/Python/executor_cases.c.h b/Python/executor_cases.c.h
diff --git a/Python/generated_cases.c.h b/Python/generated_cases.c.h
diff --git a/Python/optimizer_analysis.c b/Python/optimizer_analysis.c
@@ -345,7 +345,7 @@ remove_globals(_PyInterpreterFrame *frame, _PyUOpInstruction *buffer,
 #define sym_new_tuple _Py_uop_sym_new_tuple
 #define sym_tuple_getitem _Py_uop_sym_tuple_getitem
 #define sym_tuple_length _Py_uop_sym_tuple_length
-#define sym_is_immortal _Py_uop_sym_is_immortal
+#define sym_is_immortal _Py_uop_symbol_is_immortal
 #define sym_is_compact_int _Py_uop_sym_is_compact_int
 #define sym_new_compact_int _Py_uop_sym_new_compact_int
 #define sym_new_truthiness _Py_uop_sym_new_truthiness

@@ -34,7 +34,7 @@ typedef struct _Py_UOpsAbstractFrame _Py_UOpsAbstractFrame;
 #define sym_new_tuple _Py_uop_sym_new_tuple
 #define sym_tuple_getitem _Py_uop_sym_tuple_getitem
 #define sym_tuple_length _Py_uop_sym_tuple_length
-#define sym_is_immortal _Py_uop_sym_is_immortal
+#define sym_is_immortal _Py_uop_symbol_is_immortal
 #define sym_new_compact_int _Py_uop_sym_new_compact_int
 #define sym_is_compact_int _Py_uop_sym_is_compact_int
 #define sym_new_truthiness _Py_uop_sym_new_truthiness
@@ -318,7 +318,7 @@ dummy_func(void) {
         }
     }
 
-    op(_BINARY_OP_ADD_UNICODE, (left, right -- res)) {
+    op(_BINARY_OP_ADD_UNICODE, (left, right -- res, l, r)) {
         if (sym_is_const(ctx, left) && sym_is_const(ctx, right)) {
             assert(PyUnicode_CheckExact(sym_get_const(ctx, left)));
             assert(PyUnicode_CheckExact(sym_get_const(ctx, right)));
@@ -327,10 +327,14 @@ dummy_func(void) {
                 goto error;
             }
             res = sym_new_const(ctx, temp);
+            l = left;
+            r = right;
             Py_DECREF(temp);
         }
         else {
             res = sym_new_type(ctx, &PyUnicode_Type);
+            l = left;
+            r = right;
         }
     }
 
@@ -561,11 +565,39 @@ dummy_func(void) {
         value = PyJitRef_Borrow(sym_new_const(ctx, ptr));
     }
 
+    op(_POP_TOP_UNICODE, (value -- )) {
+        if (PyJitRef_IsBorrowed(value) ||
+            sym_is_immortal(PyJitRef_Unwrap(value))) {
+            REPLACE_OP(this_instr, _POP_TOP_NOP, 0, 0);
+        }
+    }
+
     op(_COPY, (bottom, unused[oparg-1] -- bottom, unused[oparg-1], top)) {
         assert(oparg > 0);
         top = bottom;
     }
 
+    op(_POP_TOP, (value -- )) {
+        PyTypeObject *typ = sym_get_type(value);
+        PyObject *const_val = sym_get_const(ctx, value);
+        if (PyJitRef_IsBorrowed(value) ||
+            (const_val != NULL && _Py_IsImmortal(const_val))) {
+            REPLACE_OP(this_instr, _POP_TOP_NOP, 0, 0);
+        }
+        else if (typ == &PyLong_Type) {
+            REPLACE_OP(this_instr, _POP_TOP_INT, 0, 0);
+        }
+        else if (typ == &PyFloat_Type) {
+            REPLACE_OP(this_instr, _POP_TOP_FLOAT, 0, 0);
+        }
+        else if (typ == &PyUnicode_Type) {
+            REPLACE_OP(this_instr, _POP_TOP_UNICODE, 0, 0);
+        }
+        else if (typ == &PyBool_Type) {
+            REPLACE_OP(this_instr, _POP_TOP_NOP, 0, 0);
+        }
+    }
+
     op(_SWAP, (bottom, unused[oparg-2], top -- bottom, unused[oparg-2], top)) {
         JitOptRef temp = bottom;
         bottom = top;
@@ -803,7 +835,7 @@ dummy_func(void) {
     }
 
     op(_RETURN_VALUE, (retval -- res)) {
-        JitOptRef temp = retval;
+        JitOptRef temp = PyJitRef_Wrap(PyJitRef_Unwrap(retval));
         DEAD(retval);
         SAVE_STACK();
         ctx->frame->stack_pointer = stack_pointer;

diff --git a/Python/optimizer_cases.c.h b/Python/optimizer_cases.c.h