May 4, 2024 · Apr 26, 2024 · Apr 26, 2024 · Apr 26, 2024 · Apr 26, 2024 · Apr 30, 2024
diff --git a/Include/internal/pycore_opcode_metadata.h b/Include/internal/pycore_opcode_metadata.h
diff --git a/Include/internal/pycore_uop_ids.h b/Include/internal/pycore_uop_ids.h
diff --git a/Include/opcode_ids.h b/Include/opcode_ids.h
diff --git a/Lib/_opcode_metadata.py b/Lib/_opcode_metadata.py
diff --git a/Python/bytecodes.c b/Python/bytecodes.c
        family(CALL, INLINE_CACHE_ENTRIES_CALL) = {
            CALL_BOUND_METHOD_EXACT_ARGS,
            CALL_PY_EXACT_ARGS,
            CALL_PY_WITH_DEFAULTS,
            CALL_TYPE_1,
            CALL_STR_1,
            CALL_TUPLE_1,
            _SAVE_RETURN_OFFSET +
            _PUSH_FRAME;

        inst(CALL_PY_WITH_DEFAULTS, (unused/1, func_version/2, callable, self_or_null, args[oparg] -- unused)) {
            DEOPT_IF(tstate->interp->eval_frame);
            int argcount = oparg;
            if (self_or_null != NULL) {
                args--;
                argcount++;
            }
            DEOPT_IF(!PyFunction_Check(callable));
            PyFunctionObject *func = (PyFunctionObject *)callable;
            DEOPT_IF(func->func_version != func_version);
            PyCodeObject *code = (PyCodeObject *)func->func_code;
            assert(func->func_defaults);
            assert(PyTuple_CheckExact(func->func_defaults));
            int defcount = (int)PyTuple_GET_SIZE(func->func_defaults);
            assert(defcount <= code->co_argcount);
            int min_args = code->co_argcount - defcount;
            DEOPT_IF(argcount > code->co_argcount);
            DEOPT_IF(argcount < min_args);
            DEOPT_IF(!_PyThreadState_HasStackSpace(tstate, code->co_framesize));
            STAT_INC(CALL, hit);
            _PyInterpreterFrame *new_frame = _PyFrame_PushUnchecked(tstate, func, code->co_argcount);
            for (int i = 0; i < argcount; i++) {
                new_frame->localsplus[i] = args[i];
            }
            for (int i = argcount; i < code->co_argcount; i++) {
                PyObject *def = PyTuple_GET_ITEM(func->func_defaults, i - min_args);
                new_frame->localsplus[i] = Py_NewRef(def);
            }
            // Manipulate stack and cache directly since we leave using DISPATCH_INLINED().
            STACK_SHRINK(oparg + 2);
            frame->return_offset = (uint16_t)(next_instr - this_instr);
            DISPATCH_INLINED(new_frame);
        }

        inst(CALL_TYPE_1, (unused/1, unused/2, callable, null, arg -- res)) {
            assert(oparg == 1);
            DEOPT_IF(null != NULL);
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/opcode_targets.h b/Python/opcode_targets.h
diff --git a/Python/optimizer_cases.c.h b/Python/optimizer_cases.c.h
Original file line number	Diff line number	Diff line change
Expand Up		@@ -3041,7 +3041,6 @@ dummy_func(
		family(CALL, INLINE_CACHE_ENTRIES_CALL) = {
		CALL_BOUND_METHOD_EXACT_ARGS,
		CALL_PY_EXACT_ARGS,
		CALL_PY_WITH_DEFAULTS,
		CALL_TYPE_1,
		CALL_STR_1,
		CALL_TUPLE_1,
Expand DownExpand Up		@@ -3323,40 +3322,6 @@ dummy_func(
		_SAVE_RETURN_OFFSET +
		_PUSH_FRAME;

		inst(CALL_PY_WITH_DEFAULTS, (unused/1, func_version/2, callable, self_or_null, args[oparg] -- unused)) {
Copy link Member brandtbucherMay 3, 2024 Choose a reason for hiding this comment The reason will be displayed to describe this comment to others.Learn more. Just curious, this isn't worth specializing anymore? I forget exactly why it can't be converted to tier two (I'm guessing that there's more than one reason, since the use of`this_instr` can be easily worked around). Copy link MemberAuthor markshannonMay 4, 2024 Choose a reason for hiding this comment The reason will be displayed to describe this comment to others.Learn more. Two reasons: Because this loops over both arguments and defaults, it probably isn't much faster than`CALL_PY_GENERAL` for tier 1. It is almost certainly slower in the JIT because of its size.`CALL_PY_GENERAL` calls a help function, so is more compact. We expect that, for 3.14, the tier 2 optimizer will convert them both to the same optimal sequence of operations.
		DEOPT_IF(tstate->interp->eval_frame);
		int argcount = oparg;
		if (self_or_null != NULL) {
		args--;
		argcount++;
		}
		DEOPT_IF(!PyFunction_Check(callable));
		PyFunctionObject func = (PyFunctionObject )callable;
		DEOPT_IF(func->func_version != func_version);
		PyCodeObject code = (PyCodeObject )func->func_code;
		assert(func->func_defaults);
		assert(PyTuple_CheckExact(func->func_defaults));
		int defcount = (int)PyTuple_GET_SIZE(func->func_defaults);
		assert(defcount <= code->co_argcount);
		int min_args = code->co_argcount - defcount;
		DEOPT_IF(argcount > code->co_argcount);
		DEOPT_IF(argcount < min_args);
		DEOPT_IF(!_PyThreadState_HasStackSpace(tstate, code->co_framesize));
		STAT_INC(CALL, hit);
		_PyInterpreterFrame *new_frame = _PyFrame_PushUnchecked(tstate, func, code->co_argcount);
		for (int i = 0; i < argcount; i++) {
		new_frame->localsplus[i] = args[i];
		}
		for (int i = argcount; i < code->co_argcount; i++) {
		PyObject *def = PyTuple_GET_ITEM(func->func_defaults, i - min_args);
		new_frame->localsplus[i] = Py_NewRef(def);
		}
		// Manipulate stack and cache directly since we leave using DISPATCH_INLINED().
		STACK_SHRINK(oparg + 2);
		frame->return_offset = (uint16_t)(next_instr - this_instr);
		DISPATCH_INLINED(new_frame);
		}

		inst(CALL_TYPE_1, (unused/1, unused/2, callable, null, arg -- res)) {
		assert(oparg == 1);
		DEOPT_IF(null != NULL);
Expand Down