MATLAB中数值型数据的机内编码与存贮空间浪费

Internal code of computer in MATLAB and waste of storage space

MATLAB中的数值型数据分为整数型和浮点型,整数型数据以二进制补码存储,浮点型数据存储复杂。针对此情况,在MATLAB 2012a环境中,对浮点型单精度实数和双精度实数的机内编码方案进行实验研究,结果表明浮点型数据表示为规范形式,存储其中的数符码、指数码和尾数编码。根据实数绝对值大小范围的不同,尾数有两种不同的机内编码方案:"大实数"尾码和"小实数"尾码。MATLAB浮点型数据的机内编码一方面对所表示数的范围做了扩充,另一方面实数编码方案存在大量存贮空间闲置。为充分利用存储资源,MATLAB软件有必要进一步优化。

Numeric data in MATLAB is divided into integral and Floating-point type, and integral type is stored in binary complement, but floating-Point data storage is very complicated. For this situation, in the MATLAB 2012a version, a simulated experiment was conducted to in terms of the single precision real number and double precision real number coding scheme within the machine. The experimental results showed that floating-point number is specified for standard form, and this form consisted of the storage of number sign code, index number and mantissa code. According to differences in size range of absolute value of real numbers, mantissa code has two coding scheme, and they are ＂big real number＂ and ＂small real number＂. The results found that on the one hand the range of machine number is expanded in MATLAB floating-point data coding scheme. On the other hand, coding scheme allows for plenty of idle storage space. To make full use of storage resources, it is necessary to further optimize the MATLAB software.