Is this conditional execution scheme actually used in multi ALU cores given that it has such bad worst case performance? How feasible is it to redistribute work amongst ALUs based on which conditional the stream falls into?