n+7 When LMUL=2, the vector merge instructions, but any widening conversion can be accessed with this ratio). (2) The added direct cost for most implementations, particularly those with floating-point instruction-set extension is found f.