IEEE754标准的浮点数存储格式-白红宇

IEEE754标准的浮点数存储格式

阅读量：6280 次

发布时间：2019-06-22

本文共 6990 字，大约阅读时间需要 23 分钟。

操作系统： CentOS7.3.1611_x64

gcc版本：4.8.5

基本存储格式（从高到低）： Sign + Exponent + Fraction

Sign ：符号位

Exponent ：阶码

Fraction ：有效数字

32位浮点数存储格式解析

Sign ： 1 bit（第31个bit）

Exponent ：8 bits （第 30 至 23 共 8 个bits）

Fraction ：23 bits （第 22 至 0 共 23 个bits）

32位非0浮点数的真值为(python语法) ：

(-1) **Sign * 2 **(Exponent-127) * (1 + Fraction)

示例如下：

a = 12.5

1、求解符号位

a大于0，则 Sign 为 0 ，用二进制表示为： 0

2、求解阶码

a表示为二进制为： 1100.0

小数点需要向左移动3位，则 Exponent 为 130 （127 + 3），用二进制表示为： 10000010

3、求解有效数字

有效数字需要去掉最高位隐含的1，则有效数字的整数部分为： 100

将十进制的小数转换为二进制的小数的方法为将小数*2，取整数部分，则小数部分为： 1

后面补0，则a的二进制可表示为： 01000001010010000000000000000000

即： 0100 0001 0100 1000 0000 0000 0000 0000

用16进制表示： 0x41480000

4、还原真值

Sign = bin(0) = 0Exponent = bin(10000010) = 130Fraction = bin(0.1001) = 2 ** (-1) + 2 ** (-4) = 0.5625

真值：

(-1) **0 * 2 **(130-127) * (1 + 0.5625) = 12.5

32位浮点数二进制存储解析代码（c++）：

运行效果：

[root@localhost floatTest1]# ./floatToBin1sizeof(float) : 4sizeof(int) : 4a = 12.500000showFloat : 0x 41 48 00 00UFP : 0,82,480000b : 0x41480000showIEEE754 a = 12.500000showIEEE754 varTmp = 0x00c00000showIEEE754 c = 0x00400000showIEEE754 i = 19 , a1 = 1.000000 , showIEEE754 c = 00480000 , showIEEE754 b = 0x41000000showIEEE754 i = 18 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 17 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 16 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 15 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 14 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 13 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 12 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 11 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 10 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 9 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 8 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 7 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 6 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 5 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 4 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 3 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 2 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 i = 1 , a1 = 0.000000 , showIEEE754 b = 0x41000000showIEEE754 : 0x41480000[root@localhost floatTest1]#

64位浮点数存储格式解析

Sign ： 1 bit（第31个bit）

Exponent ：11 bits （第 62 至 52 共 11 个bits）

Fraction ：52 bits （第 51 至 0 共 52 个bits）

64位非0浮点数的真值为(python语法) ：

(-1) **Sign * 2 **(Exponent-1023) * (1 + Fraction)

示例如下：

a = 12.5

1、求解符号位

a大于0，则 Sign 为 0 ，用二进制表示为： 0

2、求解阶码

a表示为二进制为： 1100.0

小数点需要向左移动3位，则 Exponent 为 1026 （1023 + 3），用二进制表示为： 10000000010

3、求解有效数字

有效数字需要去掉最高位隐含的1，则有效数字的整数部分为： 100

将十进制的小数转换为二进制的小数的方法为将小数*2，取整数部分，则小数部分为： 1

后面补0，则a的二进制可表示为：

0100000000101001000000000000000000000000000000000000000000000000

即： 0100 0000 0010 1001 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000

用16进制表示： 0x4029000000000000

4、还原真值

Sign = bin(0) = 0Exponent = bin(10000000010) = 1026Fraction = bin(0.1001) = 2 ** (-1) + 2 ** (-4) = 0.5625

真值：

(-1) **0 * 2 **(1026-1023) * (1 + 0.5625) = 12.5

64位浮点数二进制存储解析代码（c++）：

运行效果：

[root@localhost t1]# ./doubleToBin1sizeof(double) : 8sizeof(long) : 8a = 12.500000showDouble : 0x 40 29 00 00 00 00 00 00UFP : 0,402,0b : 0x0showIEEE754 a = 12.500000showIEEE754 logLen = 3showIEEE754 c = 4620693217682128896(0x4020000000000000)showIEEE754 b = 0x4020000000000000showIEEE754 varTmp = 0x8000000000000showIEEE754 c = 0x8000000000000showIEEE754 i = 48 , a1 = 1.000000 , showIEEE754 c = 9000000000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 47 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 46 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 45 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 44 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 43 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 42 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 41 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 40 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 39 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 38 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 37 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 36 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 35 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 34 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 33 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 32 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 31 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 30 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 29 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 28 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 27 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 26 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 25 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 24 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 23 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 22 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 21 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 20 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 19 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 18 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 17 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 16 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 15 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 14 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 13 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 12 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 11 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 10 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 9 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 8 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 7 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 6 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 5 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 4 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 3 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 2 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 i = 1 , a1 = 0.000000 , showIEEE754 b = 0x4020000000000000showIEEE754 : 0x4029000000000000[root@localhost t1]#

好，就这些了，希望对你有帮助。

本文github地址：

欢迎补充

转载地址：http://vdiva.baihongyu.com/

你可能感兴趣的文章

外媒：全球科技进入中美两强时代

查看>>

美国国家安全局在英国运营数据中心从事间谍活动

查看>>

【CVPR 2021主席出炉】谭铁牛、虞晶怡当选，未来4年6位华人主席

查看>>

CRM给企业带来的五大成效