From 458d82fab1d726f9ab361eda6f7987111fc6d79a Mon Sep 17 00:00:00 2001 From: Mark Shannon Date: Fri, 19 Apr 2024 16:48:04 +0100 Subject: [PATCH] Covert DEOPT_IFs on likely side exits to EXIT_IFs --- Include/internal/pycore_opcode_metadata.h | 10 +++++----- Include/internal/pycore_uop_metadata.h | 14 +++++++------- Python/bytecodes.c | 22 +++++++++++----------- Python/optimizer.c | 2 +- 4 files changed, 24 insertions(+), 24 deletions(-) diff --git a/Include/internal/pycore_opcode_metadata.h b/Include/internal/pycore_opcode_metadata.h index e75de9d12e0c55..5636debbf4a7f2 100644 --- a/Include/internal/pycore_opcode_metadata.h +++ b/Include/internal/pycore_opcode_metadata.h @@ -984,7 +984,7 @@ const struct opcode_metadata _PyOpcode_opcode_metadata[268] = { [CACHE] = { true, INSTR_FMT_IX, HAS_ESCAPES_FLAG }, [CALL] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_ERROR_FLAG | HAS_ERROR_NO_POP_FLAG | HAS_ESCAPES_FLAG }, [CALL_ALLOC_AND_ENTER_INIT] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ERROR_NO_POP_FLAG | HAS_ESCAPES_FLAG }, - [CALL_BOUND_METHOD_EXACT_ARGS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_DEOPT_FLAG }, + [CALL_BOUND_METHOD_EXACT_ARGS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_DEOPT_FLAG | HAS_EXIT_FLAG }, [CALL_BUILTIN_CLASS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, [CALL_BUILTIN_FAST] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, [CALL_BUILTIN_FAST_WITH_KEYWORDS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, @@ -1000,7 +1000,7 @@ const struct opcode_metadata _PyOpcode_opcode_metadata[268] = { [CALL_METHOD_DESCRIPTOR_FAST_WITH_KEYWORDS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, [CALL_METHOD_DESCRIPTOR_NOARGS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, [CALL_METHOD_DESCRIPTOR_O] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, - [CALL_PY_EXACT_ARGS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_DEOPT_FLAG }, + [CALL_PY_EXACT_ARGS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_DEOPT_FLAG | HAS_EXIT_FLAG }, [CALL_PY_WITH_DEFAULTS] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_DEOPT_FLAG }, [CALL_STR_1] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, [CALL_TUPLE_1] = { true, INSTR_FMT_IBC00, HAS_ARG_FLAG | HAS_EVAL_BREAK_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, @@ -1036,9 +1036,9 @@ const struct opcode_metadata _PyOpcode_opcode_metadata[268] = { [FORMAT_WITH_SPEC] = { true, INSTR_FMT_IX, HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, [FOR_ITER] = { true, INSTR_FMT_IBC, HAS_ARG_FLAG | HAS_JUMP_FLAG | HAS_ERROR_FLAG | HAS_ERROR_NO_POP_FLAG | HAS_ESCAPES_FLAG }, [FOR_ITER_GEN] = { true, INSTR_FMT_IBC, HAS_ARG_FLAG | HAS_DEOPT_FLAG }, - [FOR_ITER_LIST] = { true, INSTR_FMT_IBC, HAS_ARG_FLAG | HAS_JUMP_FLAG | HAS_DEOPT_FLAG }, - [FOR_ITER_RANGE] = { true, INSTR_FMT_IBC, HAS_ARG_FLAG | HAS_JUMP_FLAG | HAS_DEOPT_FLAG | HAS_ERROR_FLAG }, - [FOR_ITER_TUPLE] = { true, INSTR_FMT_IBC, HAS_ARG_FLAG | HAS_JUMP_FLAG | HAS_DEOPT_FLAG }, + [FOR_ITER_LIST] = { true, INSTR_FMT_IBC, HAS_ARG_FLAG | HAS_JUMP_FLAG | HAS_EXIT_FLAG }, + [FOR_ITER_RANGE] = { true, INSTR_FMT_IBC, HAS_ARG_FLAG | HAS_JUMP_FLAG | HAS_EXIT_FLAG | HAS_ERROR_FLAG }, + [FOR_ITER_TUPLE] = { true, INSTR_FMT_IBC, HAS_ARG_FLAG | HAS_JUMP_FLAG | HAS_EXIT_FLAG }, [GET_AITER] = { true, INSTR_FMT_IX, HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, [GET_ANEXT] = { true, INSTR_FMT_IX, HAS_ERROR_FLAG | HAS_ERROR_NO_POP_FLAG | HAS_ESCAPES_FLAG }, [GET_AWAITABLE] = { true, INSTR_FMT_IB, HAS_ARG_FLAG | HAS_ERROR_FLAG | HAS_ESCAPES_FLAG }, diff --git a/Include/internal/pycore_uop_metadata.h b/Include/internal/pycore_uop_metadata.h index 44ede3e77c68e1..bebeb947e23063 100644 --- a/Include/internal/pycore_uop_metadata.h +++ b/Include/internal/pycore_uop_metadata.h @@ -167,14 +167,14 @@ const uint16_t _PyUop_Flags[MAX_UOP_ID+1] = { [_GET_ITER] = HAS_ERROR_FLAG | HAS_ESCAPES_FLAG, [_GET_YIELD_FROM_ITER] = HAS_ERROR_FLAG | HAS_ERROR_NO_POP_FLAG | HAS_ESCAPES_FLAG, [_FOR_ITER_TIER_TWO] = HAS_DEOPT_FLAG | HAS_ERROR_FLAG | HAS_ERROR_NO_POP_FLAG | HAS_ESCAPES_FLAG, - [_ITER_CHECK_LIST] = HAS_DEOPT_FLAG, - [_GUARD_NOT_EXHAUSTED_LIST] = HAS_DEOPT_FLAG, + [_ITER_CHECK_LIST] = HAS_EXIT_FLAG, + [_GUARD_NOT_EXHAUSTED_LIST] = HAS_EXIT_FLAG, [_ITER_NEXT_LIST] = 0, - [_ITER_CHECK_TUPLE] = HAS_DEOPT_FLAG, - [_GUARD_NOT_EXHAUSTED_TUPLE] = HAS_DEOPT_FLAG, + [_ITER_CHECK_TUPLE] = HAS_EXIT_FLAG, + [_GUARD_NOT_EXHAUSTED_TUPLE] = HAS_EXIT_FLAG, [_ITER_NEXT_TUPLE] = 0, - [_ITER_CHECK_RANGE] = HAS_DEOPT_FLAG, - [_GUARD_NOT_EXHAUSTED_RANGE] = HAS_DEOPT_FLAG, + [_ITER_CHECK_RANGE] = HAS_EXIT_FLAG, + [_GUARD_NOT_EXHAUSTED_RANGE] = HAS_EXIT_FLAG, [_ITER_NEXT_RANGE] = HAS_ERROR_FLAG, [_WITH_EXCEPT_START] = HAS_ERROR_FLAG | HAS_ESCAPES_FLAG, [_PUSH_EXC_INFO] = 0, @@ -190,7 +190,7 @@ const uint16_t _PyUop_Flags[MAX_UOP_ID+1] = { [_CHECK_CALL_BOUND_METHOD_EXACT_ARGS] = HAS_ARG_FLAG | HAS_DEOPT_FLAG, [_INIT_CALL_BOUND_METHOD_EXACT_ARGS] = HAS_ARG_FLAG, [_CHECK_PEP_523] = HAS_DEOPT_FLAG, - [_CHECK_FUNCTION_EXACT_ARGS] = HAS_ARG_FLAG | HAS_DEOPT_FLAG, + [_CHECK_FUNCTION_EXACT_ARGS] = HAS_ARG_FLAG | HAS_EXIT_FLAG, [_CHECK_STACK_SPACE] = HAS_ARG_FLAG | HAS_DEOPT_FLAG, [_INIT_CALL_PY_EXACT_ARGS_0] = HAS_PURE_FLAG, [_INIT_CALL_PY_EXACT_ARGS_1] = HAS_PURE_FLAG, diff --git a/Python/bytecodes.c b/Python/bytecodes.c index c34d702f06418e..caaeff8a743190 100644 --- a/Python/bytecodes.c +++ b/Python/bytecodes.c @@ -2604,7 +2604,7 @@ dummy_func( } op(_ITER_CHECK_LIST, (iter -- iter)) { - DEOPT_IF(Py_TYPE(iter) != &PyListIter_Type); + EXIT_IF(Py_TYPE(iter) != &PyListIter_Type); } replaced op(_ITER_JUMP_LIST, (iter -- iter)) { @@ -2633,8 +2633,8 @@ dummy_func( _PyListIterObject *it = (_PyListIterObject *)iter; assert(Py_TYPE(iter) == &PyListIter_Type); PyListObject *seq = it->it_seq; - DEOPT_IF(seq == NULL); - DEOPT_IF((size_t)it->it_index >= (size_t)PyList_GET_SIZE(seq)); + EXIT_IF(seq == NULL); + EXIT_IF((size_t)it->it_index >= (size_t)PyList_GET_SIZE(seq)); } op(_ITER_NEXT_LIST, (iter -- iter, next)) { @@ -2653,7 +2653,7 @@ dummy_func( _ITER_NEXT_LIST; op(_ITER_CHECK_TUPLE, (iter -- iter)) { - DEOPT_IF(Py_TYPE(iter) != &PyTupleIter_Type); + EXIT_IF(Py_TYPE(iter) != &PyTupleIter_Type); } replaced op(_ITER_JUMP_TUPLE, (iter -- iter)) { @@ -2679,8 +2679,8 @@ dummy_func( _PyTupleIterObject *it = (_PyTupleIterObject *)iter; assert(Py_TYPE(iter) == &PyTupleIter_Type); PyTupleObject *seq = it->it_seq; - DEOPT_IF(seq == NULL); - DEOPT_IF(it->it_index >= PyTuple_GET_SIZE(seq)); + EXIT_IF(seq == NULL); + EXIT_IF(it->it_index >= PyTuple_GET_SIZE(seq)); } op(_ITER_NEXT_TUPLE, (iter -- iter, next)) { @@ -2700,7 +2700,7 @@ dummy_func( op(_ITER_CHECK_RANGE, (iter -- iter)) { _PyRangeIterObject *r = (_PyRangeIterObject *)iter; - DEOPT_IF(Py_TYPE(r) != &PyRangeIter_Type); + EXIT_IF(Py_TYPE(r) != &PyRangeIter_Type); } replaced op(_ITER_JUMP_RANGE, (iter -- iter)) { @@ -2720,7 +2720,7 @@ dummy_func( op(_GUARD_NOT_EXHAUSTED_RANGE, (iter -- iter)) { _PyRangeIterObject *r = (_PyRangeIterObject *)iter; assert(Py_TYPE(r) == &PyRangeIter_Type); - DEOPT_IF(r->len <= 0); + EXIT_IF(r->len <= 0); } op(_ITER_NEXT_RANGE, (iter -- iter, next)) { @@ -3121,11 +3121,11 @@ dummy_func( } op(_CHECK_FUNCTION_EXACT_ARGS, (func_version/2, callable, self_or_null, unused[oparg] -- callable, self_or_null, unused[oparg])) { - DEOPT_IF(!PyFunction_Check(callable)); + EXIT_IF(!PyFunction_Check(callable)); PyFunctionObject *func = (PyFunctionObject *)callable; - DEOPT_IF(func->func_version != func_version); + EXIT_IF(func->func_version != func_version); PyCodeObject *code = (PyCodeObject *)func->func_code; - DEOPT_IF(code->co_argcount != oparg + (self_or_null != NULL)); + EXIT_IF(code->co_argcount != oparg + (self_or_null != NULL)); } op(_CHECK_STACK_SPACE, (callable, unused, unused[oparg] -- callable, unused, unused[oparg])) { diff --git a/Python/optimizer.c b/Python/optimizer.c index bb537c9111a51f..0b6bc7438a2415 100644 --- a/Python/optimizer.c +++ b/Python/optimizer.c @@ -689,7 +689,7 @@ translate_bytecode_to_trace( if (expansion->nuops > 0) { // Reserve space for nuops (+ _SET_IP + _EXIT_TRACE) int nuops = expansion->nuops; - RESERVE(nuops); + RESERVE(nuops + 1); /* One extra for exit */ if (expansion->uops[nuops-1].uop == _POP_FRAME) { // Check for trace stack underflow now: // We can't bail e.g. in the middle of