gh-115999: Specialize loading attributes from modules in free-threaded builds (#127711)

We use the same approach that was used for specialization of LOAD_GLOBAL in free-threaded builds:

_CHECK_ATTR_MODULE is renamed to _CHECK_ATTR_MODULE_PUSH_KEYS; it pushes the keys object for the following _LOAD_ATTR_MODULE_FROM_KEYS (nee _LOAD_ATTR_MODULE). This arrangement avoids having to recheck the keys version.

_LOAD_ATTR_MODULE is renamed to _LOAD_ATTR_MODULE_FROM_KEYS; it loads the value from the keys object pushed by the preceding _CHECK_ATTR_MODULE_PUSH_KEYS at the cached index.

3 files changed, 83 insertions, 78 deletions

diff --git a/Include/internal/pycore_opcode_metadata.h b/Include/internal/pycore_opcode_metadata.h
index 81dde66..28aa112 100644
--- a/Include/internal/pycore_opcode_metadata.h
+++ b/Include/internal/pycore_opcode_metadata.h
@@ -1492,7 +1492,7 @@ int _PyOpcode_max_stack_effect(int opcode, int oparg, int *effect)  {
             return 0;
         }
         case LOAD_ATTR_MODULE: {
-            *effect = Py_MAX(0, (oparg & 1));
+            *effect = Py_MAX(1, (oparg & 1));
             return 0;
         }
         case LOAD_ATTR_NONDESCRIPTOR_NO_DICT: {
@@ -2271,7 +2271,7 @@ _PyOpcode_macro_expansion[256] = {
     [LOAD_ATTR_METHOD_LAZY_DICT] = { .nuops = 3, .uops = { { _GUARD_TYPE_VERSION, 2, 1 }, { _CHECK_ATTR_METHOD_LAZY_DICT, 1, 3 }, { _LOAD_ATTR_METHOD_LAZY_DICT, 4, 5 } } },
     [LOAD_ATTR_METHOD_NO_DICT] = { .nuops = 2, .uops = { { _GUARD_TYPE_VERSION, 2, 1 }, { _LOAD_ATTR_METHOD_NO_DICT, 4, 5 } } },
     [LOAD_ATTR_METHOD_WITH_VALUES] = { .nuops = 4, .uops = { { _GUARD_TYPE_VERSION, 2, 1 }, { _GUARD_DORV_VALUES_INST_ATTR_FROM_DICT, 0, 0 }, { _GUARD_KEYS_VERSION, 2, 3 }, { _LOAD_ATTR_METHOD_WITH_VALUES, 4, 5 } } },
-    [LOAD_ATTR_MODULE] = { .nuops = 2, .uops = { { _CHECK_ATTR_MODULE, 2, 1 }, { _LOAD_ATTR_MODULE, 1, 3 } } },
+    [LOAD_ATTR_MODULE] = { .nuops = 2, .uops = { { _CHECK_ATTR_MODULE_PUSH_KEYS, 2, 1 }, { _LOAD_ATTR_MODULE_FROM_KEYS, 1, 3 } } },
     [LOAD_ATTR_NONDESCRIPTOR_NO_DICT] = { .nuops = 2, .uops = { { _GUARD_TYPE_VERSION, 2, 1 }, { _LOAD_ATTR_NONDESCRIPTOR_NO_DICT, 4, 5 } } },
     [LOAD_ATTR_NONDESCRIPTOR_WITH_VALUES] = { .nuops = 4, .uops = { { _GUARD_TYPE_VERSION, 2, 1 }, { _GUARD_DORV_VALUES_INST_ATTR_FROM_DICT, 0, 0 }, { _GUARD_KEYS_VERSION, 2, 3 }, { _LOAD_ATTR_NONDESCRIPTOR_WITH_VALUES, 4, 5 } } },
     [LOAD_ATTR_PROPERTY] = { .nuops = 5, .uops = { { _CHECK_PEP_523, 0, 0 }, { _GUARD_TYPE_VERSION, 2, 1 }, { _LOAD_ATTR_PROPERTY_FRAME, 4, 5 }, { _SAVE_RETURN_OFFSET, 7, 9 }, { _PUSH_FRAME, 0, 0 } } },
diff --git a/Include/internal/pycore_uop_ids.h b/Include/internal/pycore_uop_ids.h
index fab4ce6..4556358 100644
--- a/Include/internal/pycore_uop_ids.h
+++ b/Include/internal/pycore_uop_ids.h
@@ -55,7 +55,7 @@ extern "C" {
 #define _CHECK_AND_ALLOCATE_OBJECT 327
 #define _CHECK_ATTR_CLASS 328
 #define _CHECK_ATTR_METHOD_LAZY_DICT 329
-#define _CHECK_ATTR_MODULE 330
+#define _CHECK_ATTR_MODULE_PUSH_KEYS 330
 #define _CHECK_ATTR_WITH_HINT 331
 #define _CHECK_CALL_BOUND_METHOD_EXACT_ARGS 332
 #define _CHECK_EG_MATCH CHECK_EG_MATCH
@@ -186,115 +186,116 @@ extern "C" {
 #define _LOAD_ATTR_METHOD_NO_DICT 416
 #define _LOAD_ATTR_METHOD_WITH_VALUES 417
 #define _LOAD_ATTR_MODULE 418
-#define _LOAD_ATTR_NONDESCRIPTOR_NO_DICT 419
-#define _LOAD_ATTR_NONDESCRIPTOR_WITH_VALUES 420
-#define _LOAD_ATTR_PROPERTY_FRAME 421
-#define _LOAD_ATTR_SLOT 422
-#define _LOAD_ATTR_SLOT_0 423
-#define _LOAD_ATTR_SLOT_1 424
-#define _LOAD_ATTR_WITH_HINT 425
+#define _LOAD_ATTR_MODULE_FROM_KEYS 419
+#define _LOAD_ATTR_NONDESCRIPTOR_NO_DICT 420
+#define _LOAD_ATTR_NONDESCRIPTOR_WITH_VALUES 421
+#define _LOAD_ATTR_PROPERTY_FRAME 422
+#define _LOAD_ATTR_SLOT 423
+#define _LOAD_ATTR_SLOT_0 424
+#define _LOAD_ATTR_SLOT_1 425
+#define _LOAD_ATTR_WITH_HINT 426
 #define _LOAD_BUILD_CLASS LOAD_BUILD_CLASS
-#define _LOAD_BYTECODE 426
+#define _LOAD_BYTECODE 427
 #define _LOAD_COMMON_CONSTANT LOAD_COMMON_CONSTANT
 #define _LOAD_CONST LOAD_CONST
 #define _LOAD_CONST_IMMORTAL LOAD_CONST_IMMORTAL
-#define _LOAD_CONST_INLINE 427
-#define _LOAD_CONST_INLINE_BORROW 428
-#define _LOAD_CONST_INLINE_BORROW_WITH_NULL 429
-#define _LOAD_CONST_INLINE_WITH_NULL 430
+#define _LOAD_CONST_INLINE 428
+#define _LOAD_CONST_INLINE_BORROW 429
+#define _LOAD_CONST_INLINE_BORROW_WITH_NULL 430
+#define _LOAD_CONST_INLINE_WITH_NULL 431
 #define _LOAD_DEREF LOAD_DEREF
-#define _LOAD_FAST 431
-#define _LOAD_FAST_0 432
-#define _LOAD_FAST_1 433
-#define _LOAD_FAST_2 434
-#define _LOAD_FAST_3 435
-#define _LOAD_FAST_4 436
-#define _LOAD_FAST_5 437
-#define _LOAD_FAST_6 438
-#define _LOAD_FAST_7 439
+#define _LOAD_FAST 432
+#define _LOAD_FAST_0 433
+#define _LOAD_FAST_1 434
+#define _LOAD_FAST_2 435
+#define _LOAD_FAST_3 436
+#define _LOAD_FAST_4 437
+#define _LOAD_FAST_5 438
+#define _LOAD_FAST_6 439
+#define _LOAD_FAST_7 440
 #define _LOAD_FAST_AND_CLEAR LOAD_FAST_AND_CLEAR
 #define _LOAD_FAST_CHECK LOAD_FAST_CHECK
 #define _LOAD_FAST_LOAD_FAST LOAD_FAST_LOAD_FAST
 #define _LOAD_FROM_DICT_OR_DEREF LOAD_FROM_DICT_OR_DEREF
 #define _LOAD_FROM_DICT_OR_GLOBALS LOAD_FROM_DICT_OR_GLOBALS
-#define _LOAD_GLOBAL 440
-#define _LOAD_GLOBAL_BUILTINS 441
-#define _LOAD_GLOBAL_BUILTINS_FROM_KEYS 442
-#define _LOAD_GLOBAL_MODULE 443
-#define _LOAD_GLOBAL_MODULE_FROM_KEYS 444
+#define _LOAD_GLOBAL 441
+#define _LOAD_GLOBAL_BUILTINS 442
+#define _LOAD_GLOBAL_BUILTINS_FROM_KEYS 443
+#define _LOAD_GLOBAL_MODULE 444
+#define _LOAD_GLOBAL_MODULE_FROM_KEYS 445
 #define _LOAD_LOCALS LOAD_LOCALS
 #define _LOAD_NAME LOAD_NAME
-#define _LOAD_SMALL_INT 445
-#define _LOAD_SMALL_INT_0 446
-#define _LOAD_SMALL_INT_1 447
-#define _LOAD_SMALL_INT_2 448
-#define _LOAD_SMALL_INT_3 449
+#define _LOAD_SMALL_INT 446
+#define _LOAD_SMALL_INT_0 447
+#define _LOAD_SMALL_INT_1 448
+#define _LOAD_SMALL_INT_2 449
+#define _LOAD_SMALL_INT_3 450
 #define _LOAD_SPECIAL LOAD_SPECIAL
 #define _LOAD_SUPER_ATTR_ATTR LOAD_SUPER_ATTR_ATTR
 #define _LOAD_SUPER_ATTR_METHOD LOAD_SUPER_ATTR_METHOD
-#define _MAKE_CALLARGS_A_TUPLE 450
+#define _MAKE_CALLARGS_A_TUPLE 451
 #define _MAKE_CELL MAKE_CELL
 #define _MAKE_FUNCTION MAKE_FUNCTION
-#define _MAKE_WARM 451
+#define _MAKE_WARM 452
 #define _MAP_ADD MAP_ADD
 #define _MATCH_CLASS MATCH_CLASS
 #define _MATCH_KEYS MATCH_KEYS
 #define _MATCH_MAPPING MATCH_MAPPING
 #define _MATCH_SEQUENCE MATCH_SEQUENCE
-#define _MAYBE_EXPAND_METHOD 452
-#define _MAYBE_EXPAND_METHOD_KW 453
-#define _MONITOR_CALL 454
-#define _MONITOR_JUMP_BACKWARD 455
-#define _MONITOR_RESUME 456
+#define _MAYBE_EXPAND_METHOD 453
+#define _MAYBE_EXPAND_METHOD_KW 454
+#define _MONITOR_CALL 455
+#define _MONITOR_JUMP_BACKWARD 456
+#define _MONITOR_RESUME 457
 #define _NOP NOP
 #define _POP_EXCEPT POP_EXCEPT
-#define _POP_JUMP_IF_FALSE 457
-#define _POP_JUMP_IF_TRUE 458
+#define _POP_JUMP_IF_FALSE 458
+#define _POP_JUMP_IF_TRUE 459
 #define _POP_TOP POP_TOP
-#define _POP_TOP_LOAD_CONST_INLINE_BORROW 459
+#define _POP_TOP_LOAD_CONST_INLINE_BORROW 460
 #define _PUSH_EXC_INFO PUSH_EXC_INFO
-#define _PUSH_FRAME 460
+#define _PUSH_FRAME 461
 #define _PUSH_NULL PUSH_NULL
-#define _PY_FRAME_GENERAL 461
-#define _PY_FRAME_KW 462
-#define _QUICKEN_RESUME 463
-#define _REPLACE_WITH_TRUE 464
+#define _PY_FRAME_GENERAL 462
+#define _PY_FRAME_KW 463
+#define _QUICKEN_RESUME 464
+#define _REPLACE_WITH_TRUE 465
 #define _RESUME_CHECK RESUME_CHECK
 #define _RETURN_GENERATOR RETURN_GENERATOR
 #define _RETURN_VALUE RETURN_VALUE
-#define _SAVE_RETURN_OFFSET 465
-#define _SEND 466
-#define _SEND_GEN_FRAME 467
+#define _SAVE_RETURN_OFFSET 466
+#define _SEND 467
+#define _SEND_GEN_FRAME 468
 #define _SETUP_ANNOTATIONS SETUP_ANNOTATIONS
 #define _SET_ADD SET_ADD
 #define _SET_FUNCTION_ATTRIBUTE SET_FUNCTION_ATTRIBUTE
 #define _SET_UPDATE SET_UPDATE
-#define _START_EXECUTOR 468
-#define _STORE_ATTR 469
-#define _STORE_ATTR_INSTANCE_VALUE 470
-#define _STORE_ATTR_SLOT 471
-#define _STORE_ATTR_WITH_HINT 472
+#define _START_EXECUTOR 469
+#define _STORE_ATTR 470
+#define _STORE_ATTR_INSTANCE_VALUE 471
+#define _STORE_ATTR_SLOT 472
+#define _STORE_ATTR_WITH_HINT 473
 #define _STORE_DEREF STORE_DEREF
-#define _STORE_FAST 473
-#define _STORE_FAST_0 474
-#define _STORE_FAST_1 475
-#define _STORE_FAST_2 476
-#define _STORE_FAST_3 477
-#define _STORE_FAST_4 478
-#define _STORE_FAST_5 479
-#define _STORE_FAST_6 480
-#define _STORE_FAST_7 481
+#define _STORE_FAST 474
+#define _STORE_FAST_0 475
+#define _STORE_FAST_1 476
+#define _STORE_FAST_2 477
+#define _STORE_FAST_3 478
+#define _STORE_FAST_4 479
+#define _STORE_FAST_5 480
+#define _STORE_FAST_6 481
+#define _STORE_FAST_7 482
 #define _STORE_FAST_LOAD_FAST STORE_FAST_LOAD_FAST
 #define _STORE_FAST_STORE_FAST STORE_FAST_STORE_FAST
 #define _STORE_GLOBAL STORE_GLOBAL
 #define _STORE_NAME STORE_NAME
-#define _STORE_SLICE 482
-#define _STORE_SUBSCR 483
+#define _STORE_SLICE 483
+#define _STORE_SUBSCR 484
 #define _STORE_SUBSCR_DICT STORE_SUBSCR_DICT
 #define _STORE_SUBSCR_LIST_INT STORE_SUBSCR_LIST_INT
 #define _SWAP SWAP
-#define _TIER2_RESUME_CHECK 484
-#define _TO_BOOL 485
+#define _TIER2_RESUME_CHECK 485
+#define _TO_BOOL 486
 #define _TO_BOOL_BOOL TO_BOOL_BOOL
 #define _TO_BOOL_INT TO_BOOL_INT
 #define _TO_BOOL_LIST TO_BOOL_LIST
@@ -304,13 +305,13 @@ extern "C" {
 #define _UNARY_NEGATIVE UNARY_NEGATIVE
 #define _UNARY_NOT UNARY_NOT
 #define _UNPACK_EX UNPACK_EX
-#define _UNPACK_SEQUENCE 486
+#define _UNPACK_SEQUENCE 487
 #define _UNPACK_SEQUENCE_LIST UNPACK_SEQUENCE_LIST
 #define _UNPACK_SEQUENCE_TUPLE UNPACK_SEQUENCE_TUPLE
 #define _UNPACK_SEQUENCE_TWO_TUPLE UNPACK_SEQUENCE_TWO_TUPLE
 #define _WITH_EXCEPT_START WITH_EXCEPT_START
 #define _YIELD_VALUE YIELD_VALUE
-#define MAX_UOP_ID 486
+#define MAX_UOP_ID 487
 
 #ifdef __cplusplus
 }
diff --git a/Include/internal/pycore_uop_metadata.h b/Include/internal/pycore_uop_metadata.h
index 89fce19..dd775d3 100644
--- a/Include/internal/pycore_uop_metadata.h
+++ b/Include/internal/pycore_uop_metadata.h
@@ -152,8 +152,8 @@ const uint16_t _PyUop_Flags[MAX_UOP_ID+1] = {
     [_LOAD_ATTR_INSTANCE_VALUE_0] = HAS_DEOPT_FLAG,
     [_LOAD_ATTR_INSTANCE_VALUE_1] = HAS_DEOPT_FLAG,
     [_LOAD_ATTR_INSTANCE_VALUE] = HAS_ARG_FLAG | HAS_DEOPT_FLAG | HAS_OPARG_AND_1_FLAG,
-    [_CHECK_ATTR_MODULE] = HAS_DEOPT_FLAG,
-    [_LOAD_ATTR_MODULE] = HAS_ARG_FLAG | HAS_DEOPT_FLAG,
+    [_CHECK_ATTR_MODULE_PUSH_KEYS] = HAS_DEOPT_FLAG,
+    [_LOAD_ATTR_MODULE_FROM_KEYS] = HAS_ARG_FLAG | HAS_DEOPT_FLAG,
     [_CHECK_ATTR_WITH_HINT] = HAS_EXIT_FLAG,
     [_LOAD_ATTR_WITH_HINT] = HAS_ARG_FLAG | HAS_NAME_FLAG | HAS_DEOPT_FLAG,
     [_LOAD_ATTR_SLOT_0] = HAS_DEOPT_FLAG,
@@ -283,6 +283,7 @@ const uint16_t _PyUop_Flags[MAX_UOP_ID+1] = {
     [_CHECK_FUNCTION] = HAS_DEOPT_FLAG,
     [_LOAD_GLOBAL_MODULE] = HAS_ARG_FLAG | HAS_DEOPT_FLAG,
     [_LOAD_GLOBAL_BUILTINS] = HAS_ARG_FLAG | HAS_DEOPT_FLAG,
+    [_LOAD_ATTR_MODULE] = HAS_ARG_FLAG | HAS_DEOPT_FLAG,
     [_INTERNAL_INCREMENT_OPT_COUNTER] = 0,
     [_DYNAMIC_EXIT] = HAS_ESCAPES_FLAG,
     [_START_EXECUTOR] = 0,
@@ -346,7 +347,7 @@ const char *const _PyOpcode_uop_name[MAX_UOP_ID+1] = {
     [_CHECK_AND_ALLOCATE_OBJECT] = "_CHECK_AND_ALLOCATE_OBJECT",
     [_CHECK_ATTR_CLASS] = "_CHECK_ATTR_CLASS",
     [_CHECK_ATTR_METHOD_LAZY_DICT] = "_CHECK_ATTR_METHOD_LAZY_DICT",
-    [_CHECK_ATTR_MODULE] = "_CHECK_ATTR_MODULE",
+    [_CHECK_ATTR_MODULE_PUSH_KEYS] = "_CHECK_ATTR_MODULE_PUSH_KEYS",
     [_CHECK_ATTR_WITH_HINT] = "_CHECK_ATTR_WITH_HINT",
     [_CHECK_CALL_BOUND_METHOD_EXACT_ARGS] = "_CHECK_CALL_BOUND_METHOD_EXACT_ARGS",
     [_CHECK_EG_MATCH] = "_CHECK_EG_MATCH",
@@ -459,6 +460,7 @@ const char *const _PyOpcode_uop_name[MAX_UOP_ID+1] = {
     [_LOAD_ATTR_METHOD_NO_DICT] = "_LOAD_ATTR_METHOD_NO_DICT",
     [_LOAD_ATTR_METHOD_WITH_VALUES] = "_LOAD_ATTR_METHOD_WITH_VALUES",
     [_LOAD_ATTR_MODULE] = "_LOAD_ATTR_MODULE",
+    [_LOAD_ATTR_MODULE_FROM_KEYS] = "_LOAD_ATTR_MODULE_FROM_KEYS",
     [_LOAD_ATTR_NONDESCRIPTOR_NO_DICT] = "_LOAD_ATTR_NONDESCRIPTOR_NO_DICT",
     [_LOAD_ATTR_NONDESCRIPTOR_WITH_VALUES] = "_LOAD_ATTR_NONDESCRIPTOR_WITH_VALUES",
     [_LOAD_ATTR_PROPERTY_FRAME] = "_LOAD_ATTR_PROPERTY_FRAME",
@@ -845,10 +847,10 @@ int _PyUop_num_popped(int opcode, int oparg)
             return 1;
         case _LOAD_ATTR_INSTANCE_VALUE:
             return 1;
-        case _CHECK_ATTR_MODULE:
+        case _CHECK_ATTR_MODULE_PUSH_KEYS:
             return 0;
-        case _LOAD_ATTR_MODULE:
-            return 1;
+        case _LOAD_ATTR_MODULE_FROM_KEYS:
+            return 2;
         case _CHECK_ATTR_WITH_HINT:
             return 0;
         case _LOAD_ATTR_WITH_HINT:
@@ -1107,6 +1109,8 @@ int _PyUop_num_popped(int opcode, int oparg)
             return 0;
         case _LOAD_GLOBAL_BUILTINS:
             return 0;
+        case _LOAD_ATTR_MODULE:
+            return 1;
         case _INTERNAL_INCREMENT_OPT_COUNTER:
             return 1;
         case _DYNAMIC_EXIT: