https://github.com/python/cpython/commit/0a387b311e617a9a614c593551d3c04a37331e53 commit: 0a387b311e617a9a614c593551d3c04a37331e53 branch: main author: Tomas R. <tomas.ro...@gmail.com> committer: brandtbucher <brandtbuc...@gmail.com> date: 2025-04-24T12:54:46-07:00 summary:
GH-131798: Split up and optimize CALL_STR_1 in the JIT (GH-132849) files: A Misc/NEWS.d/next/Core_and_Builtins/2025-04-23-20-42-55.gh-issue-131798.wVQ1Gt.rst M Include/internal/pycore_opcode_metadata.h M Include/internal/pycore_uop_ids.h M Include/internal/pycore_uop_metadata.h M Lib/test/test_capi/test_opt.py M Python/bytecodes.c M Python/executor_cases.c.h M Python/generated_cases.c.h M Python/optimizer_bytecodes.c M Python/optimizer_cases.c.h diff --git a/Include/internal/pycore_opcode_metadata.h b/Include/internal/pycore_opcode_metadata.h index 55fbf4ea884f91..e276d109c6eb0d 100644 --- a/Include/internal/pycore_opcode_metadata.h +++ b/Include/internal/pycore_opcode_metadata.h @@ -1352,7 +1352,7 @@ _PyOpcode_macro_expansion[256] = { [CALL_NON_PY_GENERAL] = { .nuops = 3, .uops = { { _CHECK_IS_NOT_PY_CALLABLE, OPARG_SIMPLE, 3 }, { _CALL_NON_PY_GENERAL, OPARG_SIMPLE, 3 }, { _CHECK_PERIODIC, OPARG_SIMPLE, 3 } } }, [CALL_PY_EXACT_ARGS] = { .nuops = 7, .uops = { { _CHECK_PEP_523, OPARG_SIMPLE, 1 }, { _CHECK_FUNCTION_VERSION, 2, 1 }, { _CHECK_FUNCTION_EXACT_ARGS, OPARG_SIMPLE, 3 }, { _CHECK_STACK_SPACE, OPARG_SIMPLE, 3 }, { _INIT_CALL_PY_EXACT_ARGS, OPARG_SIMPLE, 3 }, { _SAVE_RETURN_OFFSET, OPARG_SAVE_RETURN_OFFSET, 3 }, { _PUSH_FRAME, OPARG_SIMPLE, 3 } } }, [CALL_PY_GENERAL] = { .nuops = 5, .uops = { { _CHECK_PEP_523, OPARG_SIMPLE, 1 }, { _CHECK_FUNCTION_VERSION, 2, 1 }, { _PY_FRAME_GENERAL, OPARG_SIMPLE, 3 }, { _SAVE_RETURN_OFFSET, OPARG_SAVE_RETURN_OFFSET, 3 }, { _PUSH_FRAME, OPARG_SIMPLE, 3 } } }, - [CALL_STR_1] = { .nuops = 2, .uops = { { _CALL_STR_1, OPARG_SIMPLE, 3 }, { _CHECK_PERIODIC, OPARG_SIMPLE, 3 } } }, + [CALL_STR_1] = { .nuops = 4, .uops = { { _GUARD_NOS_NULL, OPARG_SIMPLE, 3 }, { _GUARD_CALLABLE_STR_1, OPARG_SIMPLE, 3 }, { _CALL_STR_1, OPARG_SIMPLE, 3 }, { _CHECK_PERIODIC, OPARG_SIMPLE, 3 } } }, [CALL_TUPLE_1] = { .nuops = 2, .uops = { { _CALL_TUPLE_1, OPARG_SIMPLE, 3 }, { _CHECK_PERIODIC, OPARG_SIMPLE, 3 } } }, [CALL_TYPE_1] = { .nuops = 3, .uops = { { _GUARD_NOS_NULL, OPARG_SIMPLE, 3 }, { _GUARD_CALLABLE_TYPE_1, OPARG_SIMPLE, 3 }, { _CALL_TYPE_1, OPARG_SIMPLE, 3 } } }, [CHECK_EG_MATCH] = { .nuops = 1, .uops = { { _CHECK_EG_MATCH, OPARG_SIMPLE, 0 } } }, diff --git a/Include/internal/pycore_uop_ids.h b/Include/internal/pycore_uop_ids.h index 2157c6641ef58c..98b8373ee2c649 100644 --- a/Include/internal/pycore_uop_ids.h +++ b/Include/internal/pycore_uop_ids.h @@ -116,44 +116,45 @@ extern "C" { #define _GET_LEN GET_LEN #define _GET_YIELD_FROM_ITER GET_YIELD_FROM_ITER #define _GUARD_BINARY_OP_EXTEND 371 -#define _GUARD_CALLABLE_TYPE_1 372 -#define _GUARD_DORV_NO_DICT 373 -#define _GUARD_DORV_VALUES_INST_ATTR_FROM_DICT 374 -#define _GUARD_GLOBALS_VERSION 375 -#define _GUARD_IS_FALSE_POP 376 -#define _GUARD_IS_NONE_POP 377 -#define _GUARD_IS_NOT_NONE_POP 378 -#define _GUARD_IS_TRUE_POP 379 -#define _GUARD_KEYS_VERSION 380 -#define _GUARD_NOS_DICT 381 -#define _GUARD_NOS_FLOAT 382 -#define _GUARD_NOS_INT 383 -#define _GUARD_NOS_LIST 384 -#define _GUARD_NOS_NULL 385 -#define _GUARD_NOS_TUPLE 386 -#define _GUARD_NOS_UNICODE 387 -#define _GUARD_NOT_EXHAUSTED_LIST 388 -#define _GUARD_NOT_EXHAUSTED_RANGE 389 -#define _GUARD_NOT_EXHAUSTED_TUPLE 390 -#define _GUARD_TOS_ANY_SET 391 -#define _GUARD_TOS_DICT 392 -#define _GUARD_TOS_FLOAT 393 -#define _GUARD_TOS_INT 394 -#define _GUARD_TOS_LIST 395 -#define _GUARD_TOS_TUPLE 396 -#define _GUARD_TOS_UNICODE 397 -#define _GUARD_TYPE_VERSION 398 -#define _GUARD_TYPE_VERSION_AND_LOCK 399 +#define _GUARD_CALLABLE_STR_1 372 +#define _GUARD_CALLABLE_TYPE_1 373 +#define _GUARD_DORV_NO_DICT 374 +#define _GUARD_DORV_VALUES_INST_ATTR_FROM_DICT 375 +#define _GUARD_GLOBALS_VERSION 376 +#define _GUARD_IS_FALSE_POP 377 +#define _GUARD_IS_NONE_POP 378 +#define _GUARD_IS_NOT_NONE_POP 379 +#define _GUARD_IS_TRUE_POP 380 +#define _GUARD_KEYS_VERSION 381 +#define _GUARD_NOS_DICT 382 +#define _GUARD_NOS_FLOAT 383 +#define _GUARD_NOS_INT 384 +#define _GUARD_NOS_LIST 385 +#define _GUARD_NOS_NULL 386 +#define _GUARD_NOS_TUPLE 387 +#define _GUARD_NOS_UNICODE 388 +#define _GUARD_NOT_EXHAUSTED_LIST 389 +#define _GUARD_NOT_EXHAUSTED_RANGE 390 +#define _GUARD_NOT_EXHAUSTED_TUPLE 391 +#define _GUARD_TOS_ANY_SET 392 +#define _GUARD_TOS_DICT 393 +#define _GUARD_TOS_FLOAT 394 +#define _GUARD_TOS_INT 395 +#define _GUARD_TOS_LIST 396 +#define _GUARD_TOS_TUPLE 397 +#define _GUARD_TOS_UNICODE 398 +#define _GUARD_TYPE_VERSION 399 +#define _GUARD_TYPE_VERSION_AND_LOCK 400 #define _IMPORT_FROM IMPORT_FROM #define _IMPORT_NAME IMPORT_NAME -#define _INIT_CALL_BOUND_METHOD_EXACT_ARGS 400 -#define _INIT_CALL_PY_EXACT_ARGS 401 -#define _INIT_CALL_PY_EXACT_ARGS_0 402 -#define _INIT_CALL_PY_EXACT_ARGS_1 403 -#define _INIT_CALL_PY_EXACT_ARGS_2 404 -#define _INIT_CALL_PY_EXACT_ARGS_3 405 -#define _INIT_CALL_PY_EXACT_ARGS_4 406 -#define _INSERT_NULL 407 +#define _INIT_CALL_BOUND_METHOD_EXACT_ARGS 401 +#define _INIT_CALL_PY_EXACT_ARGS 402 +#define _INIT_CALL_PY_EXACT_ARGS_0 403 +#define _INIT_CALL_PY_EXACT_ARGS_1 404 +#define _INIT_CALL_PY_EXACT_ARGS_2 405 +#define _INIT_CALL_PY_EXACT_ARGS_3 406 +#define _INIT_CALL_PY_EXACT_ARGS_4 407 +#define _INSERT_NULL 408 #define _INSTRUMENTED_FOR_ITER INSTRUMENTED_FOR_ITER #define _INSTRUMENTED_INSTRUCTION INSTRUMENTED_INSTRUCTION #define _INSTRUMENTED_JUMP_FORWARD INSTRUMENTED_JUMP_FORWARD @@ -163,163 +164,163 @@ extern "C" { #define _INSTRUMENTED_POP_JUMP_IF_NONE INSTRUMENTED_POP_JUMP_IF_NONE #define _INSTRUMENTED_POP_JUMP_IF_NOT_NONE INSTRUMENTED_POP_JUMP_IF_NOT_NONE #define _INSTRUMENTED_POP_JUMP_IF_TRUE INSTRUMENTED_POP_JUMP_IF_TRUE -#define _IS_NONE 408 +#define _IS_NONE 409 #define _IS_OP IS_OP -#define _ITER_CHECK_LIST 409 -#define _ITER_CHECK_RANGE 410 -#define _ITER_CHECK_TUPLE 411 -#define _ITER_JUMP_LIST 412 -#define _ITER_JUMP_RANGE 413 -#define _ITER_JUMP_TUPLE 414 -#define _ITER_NEXT_LIST 415 -#define _ITER_NEXT_LIST_TIER_TWO 416 -#define _ITER_NEXT_RANGE 417 -#define _ITER_NEXT_TUPLE 418 -#define _JUMP_TO_TOP 419 +#define _ITER_CHECK_LIST 410 +#define _ITER_CHECK_RANGE 411 +#define _ITER_CHECK_TUPLE 412 +#define _ITER_JUMP_LIST 413 +#define _ITER_JUMP_RANGE 414 +#define _ITER_JUMP_TUPLE 415 +#define _ITER_NEXT_LIST 416 +#define _ITER_NEXT_LIST_TIER_TWO 417 +#define _ITER_NEXT_RANGE 418 +#define _ITER_NEXT_TUPLE 419 +#define _JUMP_TO_TOP 420 #define _LIST_APPEND LIST_APPEND #define _LIST_EXTEND LIST_EXTEND -#define _LOAD_ATTR 420 -#define _LOAD_ATTR_CLASS 421 +#define _LOAD_ATTR 421 +#define _LOAD_ATTR_CLASS 422 #define _LOAD_ATTR_GETATTRIBUTE_OVERRIDDEN LOAD_ATTR_GETATTRIBUTE_OVERRIDDEN -#define _LOAD_ATTR_INSTANCE_VALUE 422 -#define _LOAD_ATTR_METHOD_LAZY_DICT 423 -#define _LOAD_ATTR_METHOD_NO_DICT 424 -#define _LOAD_ATTR_METHOD_WITH_VALUES 425 -#define _LOAD_ATTR_MODULE 426 -#define _LOAD_ATTR_NONDESCRIPTOR_NO_DICT 427 -#define _LOAD_ATTR_NONDESCRIPTOR_WITH_VALUES 428 -#define _LOAD_ATTR_PROPERTY_FRAME 429 -#define _LOAD_ATTR_SLOT 430 -#define _LOAD_ATTR_WITH_HINT 431 +#define _LOAD_ATTR_INSTANCE_VALUE 423 +#define _LOAD_ATTR_METHOD_LAZY_DICT 424 +#define _LOAD_ATTR_METHOD_NO_DICT 425 +#define _LOAD_ATTR_METHOD_WITH_VALUES 426 +#define _LOAD_ATTR_MODULE 427 +#define _LOAD_ATTR_NONDESCRIPTOR_NO_DICT 428 +#define _LOAD_ATTR_NONDESCRIPTOR_WITH_VALUES 429 +#define _LOAD_ATTR_PROPERTY_FRAME 430 +#define _LOAD_ATTR_SLOT 431 +#define _LOAD_ATTR_WITH_HINT 432 #define _LOAD_BUILD_CLASS LOAD_BUILD_CLASS -#define _LOAD_BYTECODE 432 +#define _LOAD_BYTECODE 433 #define _LOAD_COMMON_CONSTANT LOAD_COMMON_CONSTANT #define _LOAD_CONST LOAD_CONST #define _LOAD_CONST_IMMORTAL LOAD_CONST_IMMORTAL -#define _LOAD_CONST_INLINE 433 -#define _LOAD_CONST_INLINE_BORROW 434 +#define _LOAD_CONST_INLINE 434 +#define _LOAD_CONST_INLINE_BORROW 435 #define _LOAD_CONST_MORTAL LOAD_CONST_MORTAL #define _LOAD_DEREF LOAD_DEREF -#define _LOAD_FAST 435 -#define _LOAD_FAST_0 436 -#define _LOAD_FAST_1 437 -#define _LOAD_FAST_2 438 -#define _LOAD_FAST_3 439 -#define _LOAD_FAST_4 440 -#define _LOAD_FAST_5 441 -#define _LOAD_FAST_6 442 -#define _LOAD_FAST_7 443 +#define _LOAD_FAST 436 +#define _LOAD_FAST_0 437 +#define _LOAD_FAST_1 438 +#define _LOAD_FAST_2 439 +#define _LOAD_FAST_3 440 +#define _LOAD_FAST_4 441 +#define _LOAD_FAST_5 442 +#define _LOAD_FAST_6 443 +#define _LOAD_FAST_7 444 #define _LOAD_FAST_AND_CLEAR LOAD_FAST_AND_CLEAR -#define _LOAD_FAST_BORROW 444 -#define _LOAD_FAST_BORROW_0 445 -#define _LOAD_FAST_BORROW_1 446 -#define _LOAD_FAST_BORROW_2 447 -#define _LOAD_FAST_BORROW_3 448 -#define _LOAD_FAST_BORROW_4 449 -#define _LOAD_FAST_BORROW_5 450 -#define _LOAD_FAST_BORROW_6 451 -#define _LOAD_FAST_BORROW_7 452 +#define _LOAD_FAST_BORROW 445 +#define _LOAD_FAST_BORROW_0 446 +#define _LOAD_FAST_BORROW_1 447 +#define _LOAD_FAST_BORROW_2 448 +#define _LOAD_FAST_BORROW_3 449 +#define _LOAD_FAST_BORROW_4 450 +#define _LOAD_FAST_BORROW_5 451 +#define _LOAD_FAST_BORROW_6 452 +#define _LOAD_FAST_BORROW_7 453 #define _LOAD_FAST_BORROW_LOAD_FAST_BORROW LOAD_FAST_BORROW_LOAD_FAST_BORROW #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 453 -#define _LOAD_GLOBAL_BUILTINS 454 -#define _LOAD_GLOBAL_MODULE 455 +#define _LOAD_GLOBAL 454 +#define _LOAD_GLOBAL_BUILTINS 455 +#define _LOAD_GLOBAL_MODULE 456 #define _LOAD_LOCALS LOAD_LOCALS #define _LOAD_NAME LOAD_NAME -#define _LOAD_SMALL_INT 456 -#define _LOAD_SMALL_INT_0 457 -#define _LOAD_SMALL_INT_1 458 -#define _LOAD_SMALL_INT_2 459 -#define _LOAD_SMALL_INT_3 460 -#define _LOAD_SPECIAL 461 +#define _LOAD_SMALL_INT 457 +#define _LOAD_SMALL_INT_0 458 +#define _LOAD_SMALL_INT_1 459 +#define _LOAD_SMALL_INT_2 460 +#define _LOAD_SMALL_INT_3 461 +#define _LOAD_SPECIAL 462 #define _LOAD_SUPER_ATTR_ATTR LOAD_SUPER_ATTR_ATTR #define _LOAD_SUPER_ATTR_METHOD LOAD_SUPER_ATTR_METHOD -#define _MAKE_CALLARGS_A_TUPLE 462 +#define _MAKE_CALLARGS_A_TUPLE 463 #define _MAKE_CELL MAKE_CELL #define _MAKE_FUNCTION MAKE_FUNCTION -#define _MAKE_WARM 463 +#define _MAKE_WARM 464 #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 464 -#define _MAYBE_EXPAND_METHOD_KW 465 -#define _MONITOR_CALL 466 -#define _MONITOR_CALL_KW 467 -#define _MONITOR_JUMP_BACKWARD 468 -#define _MONITOR_RESUME 469 +#define _MAYBE_EXPAND_METHOD 465 +#define _MAYBE_EXPAND_METHOD_KW 466 +#define _MONITOR_CALL 467 +#define _MONITOR_CALL_KW 468 +#define _MONITOR_JUMP_BACKWARD 469 +#define _MONITOR_RESUME 470 #define _NOP NOP #define _POP_EXCEPT POP_EXCEPT -#define _POP_JUMP_IF_FALSE 470 -#define _POP_JUMP_IF_TRUE 471 +#define _POP_JUMP_IF_FALSE 471 +#define _POP_JUMP_IF_TRUE 472 #define _POP_TOP POP_TOP -#define _POP_TOP_LOAD_CONST_INLINE 472 -#define _POP_TOP_LOAD_CONST_INLINE_BORROW 473 -#define _POP_TWO_LOAD_CONST_INLINE_BORROW 474 +#define _POP_TOP_LOAD_CONST_INLINE 473 +#define _POP_TOP_LOAD_CONST_INLINE_BORROW 474 +#define _POP_TWO_LOAD_CONST_INLINE_BORROW 475 #define _PUSH_EXC_INFO PUSH_EXC_INFO -#define _PUSH_FRAME 475 +#define _PUSH_FRAME 476 #define _PUSH_NULL PUSH_NULL -#define _PUSH_NULL_CONDITIONAL 476 -#define _PY_FRAME_GENERAL 477 -#define _PY_FRAME_KW 478 -#define _QUICKEN_RESUME 479 -#define _REPLACE_WITH_TRUE 480 +#define _PUSH_NULL_CONDITIONAL 477 +#define _PY_FRAME_GENERAL 478 +#define _PY_FRAME_KW 479 +#define _QUICKEN_RESUME 480 +#define _REPLACE_WITH_TRUE 481 #define _RESUME_CHECK RESUME_CHECK #define _RETURN_GENERATOR RETURN_GENERATOR #define _RETURN_VALUE RETURN_VALUE -#define _SAVE_RETURN_OFFSET 481 -#define _SEND 482 -#define _SEND_GEN_FRAME 483 +#define _SAVE_RETURN_OFFSET 482 +#define _SEND 483 +#define _SEND_GEN_FRAME 484 #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 484 -#define _STORE_ATTR 485 -#define _STORE_ATTR_INSTANCE_VALUE 486 -#define _STORE_ATTR_SLOT 487 -#define _STORE_ATTR_WITH_HINT 488 +#define _START_EXECUTOR 485 +#define _STORE_ATTR 486 +#define _STORE_ATTR_INSTANCE_VALUE 487 +#define _STORE_ATTR_SLOT 488 +#define _STORE_ATTR_WITH_HINT 489 #define _STORE_DEREF STORE_DEREF -#define _STORE_FAST 489 -#define _STORE_FAST_0 490 -#define _STORE_FAST_1 491 -#define _STORE_FAST_2 492 -#define _STORE_FAST_3 493 -#define _STORE_FAST_4 494 -#define _STORE_FAST_5 495 -#define _STORE_FAST_6 496 -#define _STORE_FAST_7 497 +#define _STORE_FAST 490 +#define _STORE_FAST_0 491 +#define _STORE_FAST_1 492 +#define _STORE_FAST_2 493 +#define _STORE_FAST_3 494 +#define _STORE_FAST_4 495 +#define _STORE_FAST_5 496 +#define _STORE_FAST_6 497 +#define _STORE_FAST_7 498 #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 498 -#define _STORE_SUBSCR 499 -#define _STORE_SUBSCR_DICT 500 -#define _STORE_SUBSCR_LIST_INT 501 +#define _STORE_SLICE 499 +#define _STORE_SUBSCR 500 +#define _STORE_SUBSCR_DICT 501 +#define _STORE_SUBSCR_LIST_INT 502 #define _SWAP SWAP -#define _TIER2_RESUME_CHECK 502 -#define _TO_BOOL 503 +#define _TIER2_RESUME_CHECK 503 +#define _TO_BOOL 504 #define _TO_BOOL_BOOL TO_BOOL_BOOL #define _TO_BOOL_INT TO_BOOL_INT -#define _TO_BOOL_LIST 504 +#define _TO_BOOL_LIST 505 #define _TO_BOOL_NONE TO_BOOL_NONE -#define _TO_BOOL_STR 505 +#define _TO_BOOL_STR 506 #define _UNARY_INVERT UNARY_INVERT #define _UNARY_NEGATIVE UNARY_NEGATIVE #define _UNARY_NOT UNARY_NOT #define _UNPACK_EX UNPACK_EX -#define _UNPACK_SEQUENCE 506 -#define _UNPACK_SEQUENCE_LIST 507 -#define _UNPACK_SEQUENCE_TUPLE 508 -#define _UNPACK_SEQUENCE_TWO_TUPLE 509 +#define _UNPACK_SEQUENCE 507 +#define _UNPACK_SEQUENCE_LIST 508 +#define _UNPACK_SEQUENCE_TUPLE 509 +#define _UNPACK_SEQUENCE_TWO_TUPLE 510 #define _WITH_EXCEPT_START WITH_EXCEPT_START #define _YIELD_VALUE YIELD_VALUE -#define MAX_UOP_ID 509 +#define MAX_UOP_ID 510 #ifdef __cplusplus } diff --git a/Include/internal/pycore_uop_metadata.h b/Include/internal/pycore_uop_metadata.h index cb99f63bfe8ffe..c0a88278fbf4f3 100644 --- a/Include/internal/pycore_uop_metadata.h +++ b/Include/internal/pycore_uop_metadata.h @@ -244,7 +244,8 @@ const uint16_t _PyUop_Flags[MAX_UOP_ID+1] = { [_GUARD_NOS_NULL] = HAS_DEOPT_FLAG, [_GUARD_CALLABLE_TYPE_1] = HAS_DEOPT_FLAG, [_CALL_TYPE_1] = HAS_ARG_FLAG | HAS_ESCAPES_FLAG, - [_CALL_STR_1] = HAS_ARG_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG, + [_GUARD_CALLABLE_STR_1] = HAS_DEOPT_FLAG, + [_CALL_STR_1] = HAS_ARG_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG, [_CALL_TUPLE_1] = HAS_ARG_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG, [_CHECK_AND_ALLOCATE_OBJECT] = HAS_ARG_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ERROR_NO_POP_FLAG | HAS_ESCAPES_FLAG, [_CREATE_INIT_FRAME] = HAS_ARG_FLAG | HAS_ERROR_FLAG | HAS_ERROR_NO_POP_FLAG | HAS_ESCAPES_FLAG, @@ -413,6 +414,7 @@ const char *const _PyOpcode_uop_name[MAX_UOP_ID+1] = { [_GET_LEN] = "_GET_LEN", [_GET_YIELD_FROM_ITER] = "_GET_YIELD_FROM_ITER", [_GUARD_BINARY_OP_EXTEND] = "_GUARD_BINARY_OP_EXTEND", + [_GUARD_CALLABLE_STR_1] = "_GUARD_CALLABLE_STR_1", [_GUARD_CALLABLE_TYPE_1] = "_GUARD_CALLABLE_TYPE_1", [_GUARD_DORV_NO_DICT] = "_GUARD_DORV_NO_DICT", [_GUARD_DORV_VALUES_INST_ATTR_FROM_DICT] = "_GUARD_DORV_VALUES_INST_ATTR_FROM_DICT", @@ -1046,6 +1048,8 @@ int _PyUop_num_popped(int opcode, int oparg) return 0; case _CALL_TYPE_1: return 3; + case _GUARD_CALLABLE_STR_1: + return 0; case _CALL_STR_1: return 3; case _CALL_TUPLE_1: diff --git a/Lib/test/test_capi/test_opt.py b/Lib/test/test_capi/test_opt.py index 7aa8feddcf26ae..67923c5bfa712e 100644 --- a/Lib/test/test_capi/test_opt.py +++ b/Lib/test/test_capi/test_opt.py @@ -1798,6 +1798,62 @@ def testfunc(n): self.assertIn("_CALL_TYPE_1", uops) self.assertNotIn("_GUARD_IS_NOT_NONE_POP", uops) + def test_call_str_1(self): + def testfunc(n): + x = 0 + for _ in range(n): + y = str(42) + if y == '42': + x += 1 + return x + + res, ex = self._run_with_optimizer(testfunc, TIER2_THRESHOLD) + self.assertEqual(res, TIER2_THRESHOLD) + self.assertIsNotNone(ex) + uops = get_opnames(ex) + self.assertIn("_CALL_STR_1", uops) + self.assertNotIn("_GUARD_NOS_NULL", uops) + self.assertNotIn("_GUARD_CALLABLE_STR_1", uops) + + def test_call_str_1_result_is_str(self): + def testfunc(n): + x = 0 + for _ in range(n): + y = str(42) + 'foo' + if y == '42foo': + x += 1 + return x + + res, ex = self._run_with_optimizer(testfunc, TIER2_THRESHOLD) + self.assertEqual(res, TIER2_THRESHOLD) + self.assertIsNotNone(ex) + uops = get_opnames(ex) + self.assertIn("_CALL_STR_1", uops) + self.assertIn("_BINARY_OP_ADD_UNICODE", uops) + self.assertNotIn("_GUARD_NOS_UNICODE", uops) + self.assertNotIn("_GUARD_TOS_UNICODE", uops) + + def test_call_str_1_result_is_const_for_str_input(self): + # Test a special case where the argument of str(arg) + # is known to be a string. The information about the + # argument being a string should be propagated to the + # result of str(arg). + def testfunc(n): + x = 0 + for _ in range(n): + y = str('foo') # string argument + if y: # _TO_BOOL_STR + _GUARD_IS_TRUE_POP are removed + x += 1 + return x + + res, ex = self._run_with_optimizer(testfunc, TIER2_THRESHOLD) + self.assertEqual(res, TIER2_THRESHOLD) + self.assertIsNotNone(ex) + uops = get_opnames(ex) + self.assertIn("_CALL_STR_1", uops) + self.assertNotIn("_TO_BOOL_STR", uops) + self.assertNotIn("_GUARD_IS_TRUE_POP", uops) + def global_identity(x): return x diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2025-04-23-20-42-55.gh-issue-131798.wVQ1Gt.rst b/Misc/NEWS.d/next/Core_and_Builtins/2025-04-23-20-42-55.gh-issue-131798.wVQ1Gt.rst new file mode 100644 index 00000000000000..71d081f7062be8 --- /dev/null +++ b/Misc/NEWS.d/next/Core_and_Builtins/2025-04-23-20-42-55.gh-issue-131798.wVQ1Gt.rst @@ -0,0 +1,2 @@ +Split ``CALL_STR_1`` into several uops allowing the JIT to remove some of +them. Patch by Tomas Roun diff --git a/Python/bytecodes.c b/Python/bytecodes.c index e988e068d9b1e8..2ad1b3217f9612 100644 --- a/Python/bytecodes.c +++ b/Python/bytecodes.c @@ -3995,17 +3995,21 @@ dummy_func( _GUARD_CALLABLE_TYPE_1 + _CALL_TYPE_1; - op(_CALL_STR_1, (callable, null, arg -- res)) { + op(_GUARD_CALLABLE_STR_1, (callable, unused, unused -- callable, unused, unused)) { PyObject *callable_o = PyStackRef_AsPyObjectBorrow(callable); + DEOPT_IF(callable_o != (PyObject *)&PyUnicode_Type); + } + + op(_CALL_STR_1, (callable, null, arg -- res)) { PyObject *arg_o = PyStackRef_AsPyObjectBorrow(arg); assert(oparg == 1); - DEOPT_IF(!PyStackRef_IsNull(null)); - DEOPT_IF(callable_o != (PyObject *)&PyUnicode_Type); STAT_INC(CALL, hit); PyObject *res_o = PyObject_Str(arg_o); DEAD(null); DEAD(callable); + (void)callable; // Silence compiler warnings about unused variables + (void)null; PyStackRef_CLOSE(arg); ERROR_IF(res_o == NULL, error); res = PyStackRef_FromPyObjectSteal(res_o); @@ -4014,6 +4018,8 @@ dummy_func( macro(CALL_STR_1) = unused/1 + unused/2 + + _GUARD_NOS_NULL + + _GUARD_CALLABLE_STR_1 + _CALL_STR_1 + _CHECK_PERIODIC; diff --git a/Python/executor_cases.c.h b/Python/executor_cases.c.h index 9664bb38ce6312..e5bba8ee959537 100644 --- a/Python/executor_cases.c.h +++ b/Python/executor_cases.c.h @@ -5172,6 +5172,17 @@ break; } + case _GUARD_CALLABLE_STR_1: { + _PyStackRef callable; + callable = stack_pointer[-3]; + PyObject *callable_o = PyStackRef_AsPyObjectBorrow(callable); + if (callable_o != (PyObject *)&PyUnicode_Type) { + UOP_STAT_INC(uopcode, miss); + JUMP_TO_JUMP_TARGET(); + } + break; + } + case _CALL_STR_1: { _PyStackRef arg; _PyStackRef null; @@ -5181,21 +5192,14 @@ arg = stack_pointer[-1]; null = stack_pointer[-2]; callable = stack_pointer[-3]; - PyObject *callable_o = PyStackRef_AsPyObjectBorrow(callable); PyObject *arg_o = PyStackRef_AsPyObjectBorrow(arg); assert(oparg == 1); - if (!PyStackRef_IsNull(null)) { - UOP_STAT_INC(uopcode, miss); - JUMP_TO_JUMP_TARGET(); - } - if (callable_o != (PyObject *)&PyUnicode_Type) { - UOP_STAT_INC(uopcode, miss); - JUMP_TO_JUMP_TARGET(); - } STAT_INC(CALL, hit); _PyFrame_SetStackPointer(frame, stack_pointer); PyObject *res_o = PyObject_Str(arg_o); stack_pointer = _PyFrame_GetStackPointer(frame); + (void)callable; + (void)null; stack_pointer += -3; assert(WITHIN_STACK_BOUNDS()); _PyFrame_SetStackPointer(frame, stack_pointer); diff --git a/Python/generated_cases.c.h b/Python/generated_cases.c.h index 961b5b25a1060b..fdd0750cf2a87d 100644 --- a/Python/generated_cases.c.h +++ b/Python/generated_cases.c.h @@ -4148,34 +4148,42 @@ next_instr += 4; INSTRUCTION_STATS(CALL_STR_1); static_assert(INLINE_CACHE_ENTRIES_CALL == 3, "incorrect cache size"); - _PyStackRef callable; _PyStackRef null; + _PyStackRef callable; _PyStackRef arg; _PyStackRef res; /* Skip 1 cache entry */ /* Skip 2 cache entries */ - // _CALL_STR_1 + // _GUARD_NOS_NULL { - arg = stack_pointer[-1]; null = stack_pointer[-2]; - callable = stack_pointer[-3]; - PyObject *callable_o = PyStackRef_AsPyObjectBorrow(callable); - PyObject *arg_o = PyStackRef_AsPyObjectBorrow(arg); - assert(oparg == 1); if (!PyStackRef_IsNull(null)) { UPDATE_MISS_STATS(CALL); assert(_PyOpcode_Deopt[opcode] == (CALL)); JUMP_TO_PREDICTED(CALL); } + } + // _GUARD_CALLABLE_STR_1 + { + callable = stack_pointer[-3]; + PyObject *callable_o = PyStackRef_AsPyObjectBorrow(callable); if (callable_o != (PyObject *)&PyUnicode_Type) { UPDATE_MISS_STATS(CALL); assert(_PyOpcode_Deopt[opcode] == (CALL)); JUMP_TO_PREDICTED(CALL); } + } + // _CALL_STR_1 + { + arg = stack_pointer[-1]; + PyObject *arg_o = PyStackRef_AsPyObjectBorrow(arg); + assert(oparg == 1); STAT_INC(CALL, hit); _PyFrame_SetStackPointer(frame, stack_pointer); PyObject *res_o = PyObject_Str(arg_o); stack_pointer = _PyFrame_GetStackPointer(frame); + (void)callable; + (void)null; stack_pointer += -3; assert(WITHIN_STACK_BOUNDS()); _PyFrame_SetStackPointer(frame, stack_pointer); diff --git a/Python/optimizer_bytecodes.c b/Python/optimizer_bytecodes.c index a14d5c037ac6f1..4f96140f8f706e 100644 --- a/Python/optimizer_bytecodes.c +++ b/Python/optimizer_bytecodes.c @@ -855,6 +855,16 @@ dummy_func(void) { } } + op(_CALL_STR_1, (unused, unused, arg -- res)) { + if (sym_matches_type(arg, &PyUnicode_Type)) { + // e.g. str('foo') or str(foo) where foo is known to be a string + res = arg; + } + else { + res = sym_new_type(ctx, &PyUnicode_Type); + } + } + op(_GUARD_IS_TRUE_POP, (flag -- )) { if (sym_is_const(ctx, flag)) { PyObject *value = sym_get_const(ctx, flag); @@ -1021,6 +1031,13 @@ dummy_func(void) { sym_set_const(callable, (PyObject *)&PyType_Type); } + op(_GUARD_CALLABLE_STR_1, (callable, unused, unused -- callable, unused, unused)) { + if (sym_get_const(ctx, callable) == (PyObject *)&PyUnicode_Type) { + REPLACE_OP(this_instr, _NOP, 0, 0); + } + sym_set_const(callable, (PyObject *)&PyUnicode_Type); + } + // END BYTECODES // } diff --git a/Python/optimizer_cases.c.h b/Python/optimizer_cases.c.h index 54a016564c1101..915dd642226892 100644 --- a/Python/optimizer_cases.c.h +++ b/Python/optimizer_cases.c.h @@ -1892,9 +1892,26 @@ break; } + case _GUARD_CALLABLE_STR_1: { + JitOptSymbol *callable; + callable = stack_pointer[-3]; + if (sym_get_const(ctx, callable) == (PyObject *)&PyUnicode_Type) { + REPLACE_OP(this_instr, _NOP, 0, 0); + } + sym_set_const(callable, (PyObject *)&PyUnicode_Type); + break; + } + case _CALL_STR_1: { + JitOptSymbol *arg; JitOptSymbol *res; - res = sym_new_not_null(ctx); + arg = stack_pointer[-1]; + if (sym_matches_type(arg, &PyUnicode_Type)) { + res = arg; + } + else { + res = sym_new_type(ctx, &PyUnicode_Type); + } stack_pointer[-3] = res; stack_pointer += -2; assert(WITHIN_STACK_BOUNDS()); _______________________________________________ Python-checkins mailing list -- python-checkins@python.org To unsubscribe send an email to python-checkins-le...@python.org https://mail.python.org/mailman3/lists/python-checkins.python.org/ Member address: arch...@mail-archive.com