chromas converstion: optimize MMX/SSE acceleration even more by leveraging out of order execution, basically avoid CPU stalls by using more registers and by spacing memory fetch and use as much as possible (CPU will run available instructions while waiting for memory fecth to complete)