作者executor1107 (口責.我受夠了)
看板NSYSU-CSE97
標題[分享] [計概] [Chapter 3.4~3.5] 隨手整理
時間Sun Dec 12 14:24:58 2004
3.4什麼時候會用到實在沒概念
有人能說一下嗎?
Representation:
To represent a number in Execess, use the following procedure:
1. Add the magic number to the integer.
2. Change the result to binary and 0s so that there is a total of N bits.
Interpretation:
To interpret a number in Execess, use the following procedure"
1. Change the number to decimal.
2. Subtract the magic number from the interger.
感覺是一種轉換過程?加密解密?...(好像還沒那麼高級吧XD)
Convert the fraction to binary:
總而言之就是乘2再取第一位就是了(或許有錯?)
0.125*2=>0.250*2=>0.500*2=>1.000
0 0 1
不過要是0.35呢?
0.35*2=>0.70=>1.4=>?
0 1
Normalize實在又臭又長
sign、exponent、matissa是要注意的
power of exponent代表的電腦能處理的數字的大小
matissa是由unsigned的型式存取的
After Normalized
Store in 2 case(Made by IEEE):
Excess_127:
sign exponent mantissa
1 8 23
1 11 52
Floating-point interpretation for single precision:
The following procedure interprets a 32-bit floating-point number stored in
memory.
1. Use the leftmost bit as the sign.
2. Change the next 8 bits to decimal and subtract 127 from it. This is the
exponent
這邊我有點問題.為啥要subtract exponent所容納的位元的signed max positive number?
像是8 bits就 subtract 127,5 bits就subtract 16?
是為了表達儲存的問題嗎?為了exponent部份不用考慮到negative的部份?
3. Add 1 and a decimal point to the next 23 bits. You can ignore any extra 0s
at the right.
4. Move the decimal point to the position using the value of exponent.
5. Change the whole part to decimal.
6. Change the fraction part to decimal.
7. Combine the whole and the fraction parts.
--
......ID......
......I am a learning Dancer......
Shake my body through the music...
Dance with the Feel...
This is my ID......
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 140.117.186.180
※ 編輯: executor1107 來自: 140.117.186.180 (12/12 14:44)