浮点运算就是实数运算，因为计算机只能存储整数，所以实数都是约数，这样浮

Floating-point operations are real-number operations, because computers can only store integers, so real numbers are all divisors, so floating-point operations are very slow and have errors. Most machines are 32-bit, that is, 32-bit is used to represent integers, then for unsigned integers is 0 to 2 ^ 32-1, for signed words is -2 ^ 31 to 2 ^ 31-1 . <br>A floating-point arithmetic unit is a structure that performs floating-point arithmetic. At this stage, it is implemented with a circuit, which is used in computer chips. It is a major development after the integer arithmetic unit. Before the invention of the floating-point arithmetic unit, the floating-point arithmetic in the computer was simulated by integer arithmetic, which was very inefficient. <br>1. The floating-point addition and subtraction calculation step <br>provided two floating point numbers X = Mx ※ 2Ex Y = My ※ 2Ey <br><br>achieve 5 X ± Y use the following steps is completed: <br>① the order of operation: a small step toward large order in line <br>② for mantissa Addition and subtraction operations <br>③ Normalization processing: the result of the mantissa operation must become a normalized floating-point number. For the two-sign complement two mantissa, it must be 001 ×××… ×× or 110 ×××… × If the form of × does not conform to the above-mentioned form, the left or right rule shall be processed. <br>④Rounding operation: When performing the order or right gauge operation, the "0" rounding "1" rounding method is often used to round the mantissa value shifted to the right to ensure accuracy. <br>⑤ determination accuracy of the results: i.e. check whether the exponent overflow <br>if underflow order code (00 ... shift is represented by 0), the result is set to 0 for the machine; <br>(maximum value exceeds the indicated order code) if the exponent underflow Set overflow flag. <br>2. Operation step 0 of floating point multiplication and division <br>① Order code operation: order code sum (multiplication) or order code difference (division) <br>, ie [Ex + Ey] shift = [Ex] shift + [Ey] complement <br>[Ex－ Ey] shift = [Ex] shift + [－Ey] complement <br>② floating point mantissa processing: the floating point number mantissa multiplication and division operation results must be rounded.

浮点运算就是实数运算，因为计算机只能存储整数，所以实数都是约数，这样浮点运算是很慢的而且会有误差。 大多数机器都是32位的，也就是说32位都用来表示整数的话，那么对于无符号整数就是0 到 2^32-1，对于有符号的话就是-2^31 到 2^31-1。<br> 浮点运算器是进行浮点运算的结构。现阶段是用电路来实现，应用在计算机芯片中。是整数运算器之后的一大发展，因为在浮点运算器发明之前，计算机中的浮点运算是都是用整数运算来模拟的，效率十分低下。<br> 1.浮点加减法的运算步骤 <br> 设两个浮点数 X=Mx※2Ex Y=My※2Ey <br><br> 实现X±Y要用如下5步完成： <br> ①对阶操作：小阶向大阶看齐 <br> ②进行尾数加减运算 <br> ③规格化处理：尾数进行运算的结果必须变成规格化的浮点数，对于双符号位的补码尾数来说，就必须是 001×××…×× 或110×××…××的形式，若不符合上述形式要进行左规或右规处理。<br> ④舍入操作：在执行对阶或右规操作时常用“0”舍“1”入法将右移出去的尾数数值进行舍入，以确保精度。 <br> ⑤判结果的正确性：即检查阶码是否溢出 <br>若阶码下溢（移码表示是00…0），要置结果为机器0； <br>若阶码上溢（超过了阶码表示的最大值）置溢出标志。<br> 2、浮点乘除法的运算步骤 0<br> ①阶码运算：阶码求和（乘法）或阶码求差（除法） <br>即 [Ex+Ey]移= [Ex]移+ [Ey]补 <br>[Ex－Ey]移= [Ex]移+ [－Ey]补<br> ②浮点数的尾数处理：浮点数中尾数乘除法运算结果要进行舍入处理 。

Floating point operation is real operation, because the computer can only store integers, so real numbers are divisors, so floating point operation is very slow and there will be errors. Most machines are 32-bit, that is to say, 32 bits are used to represent integers. For an unsigned integer, it is 0 to 2 ^ 32-1. For a signed integer, it is - 2 ^ 31 to 2 ^ 31-1.<br>Floating point arithmetic is the structure of floating-point arithmetic. At present, it is realized by circuit and applied in computer chip. It is a great development after the integer arithmetic because before the invention of the floating-point arithmetic, the floating-point arithmetic in the computer was simulated by integer arithmetic, which is very inefficient.<br>1. Operation steps of floating-point addition and subtraction<br>Set two floating-point numbers x = MX ※ 2ex y = my ※ 2ey<br>To achieve x ± y, the following five steps are required:<br>① Opposite order operation: small order to large order<br>② Add and subtract mantissa<br>③ Normalization processing: the result of mantissa operation must be a normalized floating-point number. For the complement mantissa of double sign bits, it must be 001 ×× XX or 110 XX If the form of ×× does not conform to the above form, it shall be handled with left or right gauge.<br>④ Rounding operation: in order to ensure accuracy, the rounding method of "0" to "1" is often used to round the mantissa value moved out to the right.<br>⑤ Correctness of judgment result: that is, check whether the order code overflows<br>If the order code underflow (shift code means 00 0), set the result to machine 0;<br>If the order code overflows (exceeding the maximum value represented by the order code), set the overflow flag.<br>2. Operation step 0 of floating point multiplication and division<br>① Order code operation: order code sum (multiplication) or order code difference (Division)<br>That is, [ex + ey] shift = [ex] Shift + [ey] supplement<br>[ex ey] shift = [ex] Shift + [- ey] supplement<br>② Mantissa processing of floating-point numbers: the result of mantissa multiplication and division in floating-point numbers should be rounded.<br>