Suppose we want to build a 100 Gigaflop computer (i.e. a machine capable of 1011 floating point calculations per second on some problem of interest)
There is a range of possible ways of doing this.