Move type checks from switch cases to beginning of `_Py_Specialize_BinaryOp`

[WolframAlph]

Feature or enhancement

Proposal:

_Py_Specialize_BinaryOp:

cpython/Python/specialize.c

Lines 2380 to 2443 in aeb9b65

	_Py_Specialize_BinaryOp(_PyStackRef lhs_st, _PyStackRef rhs_st, _Py_CODEUNIT *instr,
	int oparg, _PyStackRef *locals)
	{
	PyObject *lhs = PyStackRef_AsPyObjectBorrow(lhs_st);
	PyObject *rhs = PyStackRef_AsPyObjectBorrow(rhs_st);
	assert(ENABLE_SPECIALIZATION_FT);
	assert(_PyOpcode_Caches[BINARY_OP] == INLINE_CACHE_ENTRIES_BINARY_OP);
	switch (oparg) {
	case NB_ADD:
	case NB_INPLACE_ADD:
	if (!Py_IS_TYPE(lhs, Py_TYPE(rhs))) {
	break;
	}
	if (PyUnicode_CheckExact(lhs)) {
	_Py_CODEUNIT next = instr[INLINE_CACHE_ENTRIES_BINARY_OP + 1];
	bool to_store = (next.op.code == STORE_FAST);
	if (to_store && PyStackRef_AsPyObjectBorrow(locals[next.op.arg]) == lhs) {
	specialize(instr, BINARY_OP_INPLACE_ADD_UNICODE);
	return;
	}
	specialize(instr, BINARY_OP_ADD_UNICODE);
	return;
	}
	if (PyLong_CheckExact(lhs)) {
	specialize(instr, BINARY_OP_ADD_INT);
	return;
	}
	if (PyFloat_CheckExact(lhs)) {
	specialize(instr, BINARY_OP_ADD_FLOAT);
	return;
	}
	break;
	case NB_MULTIPLY:
	case NB_INPLACE_MULTIPLY:
	if (!Py_IS_TYPE(lhs, Py_TYPE(rhs))) {
	break;
	}
	if (PyLong_CheckExact(lhs)) {
	specialize(instr, BINARY_OP_MULTIPLY_INT);
	return;
	}
	if (PyFloat_CheckExact(lhs)) {
	specialize(instr, BINARY_OP_MULTIPLY_FLOAT);
	return;
	}
	break;
	case NB_SUBTRACT:
	case NB_INPLACE_SUBTRACT:
	if (!Py_IS_TYPE(lhs, Py_TYPE(rhs))) {
	break;
	}
	if (PyLong_CheckExact(lhs)) {
	specialize(instr, BINARY_OP_SUBTRACT_INT);
	return;
	}
	if (PyFloat_CheckExact(lhs)) {
	specialize(instr, BINARY_OP_SUBTRACT_FLOAT);
	return;
	}
	break;
	}
	SPECIALIZATION_FAIL(BINARY_OP, binary_op_fail_kind(oparg, lhs, rhs));
	unspecialize(instr);
	}

uses

if (!Py_IS_TYPE(lhs, Py_TYPE(rhs))) { 
    break; 
} 

check inside every switch case which is redundant. Placing this check before switch results in same performance but less generated code. Compiler with -O3 option seems to not recognize it. Attaching generated machine code (clang, apple silicon) for old and new versions
new.txt
old.txt

Feel free to close this if not worth it.

Has this already been discussed elsewhere?

No response given

Links to previous discussion of this feature:

No response

Linked PRs

• gh-128310: Move type checks from switch cases to begining of _Py_Specialize_BinaryOp #128311

[iritkatriel]

Placing this check before switch results in same performance

With your change the check will execute also when oparg is not one of the cases of the switch, so it might be less efficient.