GH-115685: Split `_TO_BOOL_ALWAYS_TRUE` into micro-ops (GH-116352)

5 files changed, 27 insertions, 21 deletions

diff --git a/Python/bytecodes.c b/Python/bytecodes.c
index 396a8f0..15794cb 100644
--- a/Python/bytecodes.c
+++ b/Python/bytecodes.c
@@ -382,15 +382,16 @@ dummy_func(
             }
         }
 
-        inst(TO_BOOL_ALWAYS_TRUE, (unused/1, version/2, value -- res)) {
-            // This one is a bit weird, because we expect *some* failures:
-            assert(version);
-            EXIT_IF(Py_TYPE(value)->tp_version_tag != version);
-            STAT_INC(TO_BOOL, hit);
-            DECREF_INPUTS();
+        op(_REPLACE_WITH_TRUE, (value -- res)) {
+            Py_DECREF(value);
             res = Py_True;
         }
 
+        macro(TO_BOOL_ALWAYS_TRUE) =
+            unused/1 +
+            _GUARD_TYPE_VERSION +
+            _REPLACE_WITH_TRUE;
+
         inst(UNARY_INVERT, (value -- res)) {
             res = PyNumber_Invert(value);
             DECREF_INPUTS();
diff --git a/Python/executor_cases.c.h b/Python/executor_cases.c.h
index 56ee938..806f748 100644
--- a/Python/executor_cases.c.h
+++ b/Python/executor_cases.c.h
@@ -383,15 +383,10 @@
             break;
         }
 
-        case _TO_BOOL_ALWAYS_TRUE: {
+        case _REPLACE_WITH_TRUE: {
             PyObject *value;
             PyObject *res;
             value = stack_pointer[-1];
-            uint32_t version = (uint32_t)CURRENT_OPERAND();
-            // This one is a bit weird, because we expect *some* failures:
-            assert(version);
-            if (Py_TYPE(value)->tp_version_tag != version) goto side_exit;
-            STAT_INC(TO_BOOL, hit);
             Py_DECREF(value);
             res = Py_True;
             stack_pointer[-1] = res;
diff --git a/Python/generated_cases.c.h b/Python/generated_cases.c.h
index e612c9e..4947c91 100644
--- a/Python/generated_cases.c.h
+++ b/Python/generated_cases.c.h
@@ -5580,17 +5580,24 @@
             next_instr += 4;
             INSTRUCTION_STATS(TO_BOOL_ALWAYS_TRUE);
             static_assert(INLINE_CACHE_ENTRIES_TO_BOOL == 3, "incorrect cache size");
+            PyObject *owner;
             PyObject *value;
             PyObject *res;
             /* Skip 1 cache entry */
-            value = stack_pointer[-1];
-            uint32_t version = read_u32(&this_instr[2].cache);
-            // This one is a bit weird, because we expect *some* failures:
-            assert(version);
-            DEOPT_IF(Py_TYPE(value)->tp_version_tag != version, TO_BOOL);
-            STAT_INC(TO_BOOL, hit);
-            Py_DECREF(value);
-            res = Py_True;
+            // _GUARD_TYPE_VERSION
+            owner = stack_pointer[-1];
+            {
+                uint32_t type_version = read_u32(&this_instr[2].cache);
+                PyTypeObject *tp = Py_TYPE(owner);
+                assert(type_version != 0);
+                DEOPT_IF(tp->tp_version_tag != type_version, TO_BOOL);
+            }
+            // _REPLACE_WITH_TRUE
+            value = owner;
+            {
+                Py_DECREF(value);
+                res = Py_True;
+            }
             stack_pointer[-1] = res;
             DISPATCH();
         }
diff --git a/Python/optimizer_analysis.c b/Python/optimizer_analysis.c
index 21cccdb..18e3382 100644
--- a/Python/optimizer_analysis.c
+++ b/Python/optimizer_analysis.c
@@ -456,6 +456,9 @@ remove_unneeded_uops(_PyUOpInstruction *buffer, int buffer_size)
                     last->opcode = _NOP;
                     buffer[pc].opcode = NOP;
                 }
+                if (last->opcode == _REPLACE_WITH_TRUE) {
+                    last->opcode = _NOP;
+                }
                 break;
             }
             case _JUMP_TO_TOP:
diff --git a/Python/optimizer_cases.c.h b/Python/optimizer_cases.c.h
index 6e65e2e..4ed1ac9 100644
--- a/Python/optimizer_cases.c.h
+++ b/Python/optimizer_cases.c.h
@@ -203,7 +203,7 @@
             break;
         }
 
-        case _TO_BOOL_ALWAYS_TRUE: {
+        case _REPLACE_WITH_TRUE: {
             _Py_UopsSymbol *res;
             res = sym_new_unknown(ctx);
             if (res == NULL) goto out_of_space;