两个16位整数相乘，乘积总是“向左增长”，这意味着多次相乘后乘积将会很快超出定点器件的数据范围。而且要将32位乘积保存到数据存储器，就要开销2个机器周期以及2个字的程序和RAM单元；并且，由于乘法器都是16位相乘，因此很难在后续的递推运算中，将32位乘积作为乘法器的输入。然而，小数相乘，乘积总是“向右增长”，这就使得超出定点器件数据范围的是我们不太感兴趣的部分。在小数乘法下，既可以存储32位乘积，也可以存储高16位乘积，这就允许用较少的资源保存结果，也便于用于递推运算中。这就是为什么定点DSP芯片都采用小数乘法的原因。-Two 16-bit integers, the product is always " left growth" , which means the multiplication product of several fixed devices will soon exceed the range of data. And you want to save the 32-bit product into the data memory, it is necessary overhead two machine cycles, and two words of program and RAM unit　And, because the multiplier is multiplied by 16, it is difficult in subsequent recursive arithmetic , the product of a multiplier 32-bit input. However, the fractional multiplication, multiplication always " right growth" , which makes the range of data beyond the fixed-point devices are not interested in our part. Under the fractional multiplication, both 32-bit product may be stored, the upper 16 bits can be stored product, which allows to save the results with fewer resources, but also easy for recursive computations. This is why the fixed-point DSP chips are used decimal multiplication causes.