Under the Hood: Data Representation. Computer Science 104 Lecture 2

Transcription

1 Under the Hood: Data Representation Computer Science 104 Lecture 2

2 Admin Piazza, Sakai Up Everyone should have access Homework 1 Posted Due Feb 6 PDF or Plain Text Only: No Word or RTF Recommended: Learn LaTeX 2

3 Last Time on CS104. Who can remind us what we talked about last time? Compsci 104 3

4 Last Time on CS104. Who can remind us what we talked about last time? Course policies Why to take this course Everything is a number it is all 1s and 0s Number representations Decimal Binary Hexadecimal ( hex ) Octal Compsci 104 4

5 Converting Numbers Hex to Binary: Expand each hex digit to 4 bits: 0xFC1 = Binary to Hex: Group up 4 bits into a digit: = 0x6B3 Binary to Decimal: Add up values of bits with 1: = = 77 5

6 Converting Numbers Converting Decimal to Binary Suppose I want to convert 4872 to binary How would I do this?. Think about it for a second 6

7 Converting Numbers Converting Decimal to Binary Suppose I want to convert 4872 to binary How would I do this?. Think about it for a second

8 Converting Numbers Converting Decimal to Binary Suppose I want to convert 4872 to binary How would I do this?. Think about it for a second

9 Converting Numbers Converting Decimal to Binary Suppose I want to convert 4872 to binary How would I do this?. Think about it for a second

10 Converting Numbers Converting Decimal to Binary Suppose I want to convert 4872 to binary How would I do this?. Think about it for a second

11 Converting Numbers Converting Decimal to Binary Suppose I want to convert 4872 to binary How would I do this?. Think about it for a second

12 What about negative numbers? This works great for N >= 0 But what about negative numbers? Sometimes we don t need them: unsigned number Sometimes kind of important: signed number Many options: Sign magnitude 1 s complement 2 s complement Biased 12

13 Sign Magnitude Use leftmost bit for + (0) or (1): 6-bit example (1 sign bit + 5 magnitude bits): +17 = = Pros: Conceptually simple Easy to convert Cons: Harder to compute (add, subtract, etc) with Positive and negative 0: and

14 1 s Complement Representation for Integers Use largest positive binary numbers to represent negative numbers To negate a number, invert ( not ) each bit: 0->1 1->0 Cons: Still two 0s Still hard to compute with

15 2 s Complement Representation for Integers To negated, flip bits, add 1: Pros: 1 s complement + 1 Easy to compute with One representation of 0 Cons: More complex negation Extra negative number (-8) Ubiquitous choice

16 Binary Math : Addition Suppose we want to add two numbers: How do we do this? 16

17 Binary Math : Addition Suppose we want to add two numbers: How do we do this? Let s revisit decimal addition Think about the process as we do it 17

18 Binary Math : Addition Suppose we want to add two numbers: First add one s digit 5+2 = 7 18

19 Binary Math : Addition Suppose we want to add two numbers: First add one s digit 5+2 = 7 Next add ten s digit 9+3 = 12 (2 carry a 1) 19

20 Binary Math : Addition Suppose we want to add two numbers: First add one s digit 5+2 = 7 Next add ten s digit 9+3 = 12 (2 carry a 1) Last add hundred s digit = 9 20

21 Binary Math : Addition Suppose we want to add two numbers: Back to the binary: First add 1 s digit 1+1 =? 21

22 Binary Math : Addition Suppose we want to add two numbers: Back to the binary: First add 1 s digit 1+1 = 2 (0 carry a 1) 22

23 Binary Math : Addition Suppose we want to add two numbers: Back to the binary: First add 1 s digit 1+1 = 2 (0 carry a 1) Then 2 s digit: =2 (0 carry a 1) You all finish it out. 23

24 Binary Math : Addition Suppose we want to add two numbers: = = = 72 Can check our work in decimal 24

25 What about this one: Binary Math : Addition

26 Binary Math : Addition What about this one: = = = -56 But that can t be right? What do you expect for the answer? What is it in 8-bit signed 2 s complement? 26

27 Integer Overflow Answer should be 200 Not representable in 8-bit signed representation No right answer Call Integer Overflow Signed addition: CI!= CO of last bit Unsigned addition: CO!= 0 of last bit Real problem in programs SNES SimCity Cheat Can easily imagine more serious cases Awareness can make you better programmer 27

28 Subtraction 2 s complement makes subtraction easy: Remember: A - B = A + (-B) And: -B = ~B + 1 that means flip bits ( not ) So we just flip the bits and start with CI = 1 Later: No new circuits to subtract (re-use add) >

29 What About Non-integer Numbers? There are infinitely many real numbers between two integers Many important numbers are real Pi = ½ = 0.5 How could we represent these sorts of numbers? Remember, we only have integers Everyone think for a few minutes Feel free to brainstorm with your neighboors 29

30 Option 1: Fixed point Use normal integers, but (X*2 K ) instead of X Example: 32 bit int, but use X* * = * = 32768, etc.. Pros: Addition/subtraction just like integers ( free ) Cons: Mul/div require renormalizing (divide by 64K) Range limited (no good rep for large + small) Can be good in specific situations 30

31 Option 2: Rational Numbers Represent each number as fraction: X / Y X and Y are integers (say 16 bits each) 0.5 = (1,2) = (31415, 1000) Pros: Represent a lot of numbers Precision in some cases (e.g., 1/3) Cons: Hard to work with Multiple representations for many numbers 31

32 Can we do better? Think about scientific notation for a second: For example: 6.02 * Real number, but comprised of ints: 6 generally only 1 digit here 2 any number here 10 always 10 (base we work in) 23 can be positive or negative Can we do something like this in binary? 32

33 Option 3: Floating Point How about: +/- X.YYYYYY * 2 +/-N Big numbers: large positive N Small numbers (<1): negative N Numbers near 0: small N This is floating point : most common way 33

34 IEEE single precision floating point Specific format called IEEE single precision: +/- 1.YYYYY * 2 (N-127) float in Java, C, C++, Assume X is always 1 (save a bit) 1 sign bit (+ = 0, 1 = -) 8 bit biased exponent (do N-127) Implicit 1 before binary point 23-bit mantissa (YYYYY) 34

35 Binary fractions 1.YYYY has a binary point Like a decimal point but in binary After a decimal point, you have tenths hundredths Thousandths. So after a binary point you have 35

36 Binary fractions 1.YYYY has a binary point Like a decimal point but in binary After a decimal point, you have Tenths Hundredths Thousandths. So after a binary point you have Halves Quarters Eights. 36

37 Floating point example Binary fraction example: = ½ + 1 / 8 = For floating point, needs normalization: * 2 2 Sign is +, which = 0 Exponent = = 129 = Mantissa =

38 Floating Point Representation Example: What floating-point number is: 0xC ? 38

39 Answer What floating-point number is 0xC ? X = s E F 0 Sign = 1 which is negtiave Exponent = (128+2)-127 = 3 Mantissa = x2 3 = =

40 Trick question How do you represent 0.0? Why is this a trick question? 40

41 Trick question How do you represent 0.0? Why is this a trick question? 0.0 = But need 1.XXXXX representation? 41

42 Trick question How do you represent 0.0? Why is this a trick question? 0.0 = But need 1.XXXXX representation? Exponent of 0 is denormalized Implicit 0. instead of 1. in mantissa Allows to be 0 Helps with very small numbers near 0 Results in +/- 0 in FP (but they are equal ) 42

43 Other weird FP numbers Exponent = also not standard All 0 mantissa: +/- 1/0 = + -1/0 = - Non zero mantissa: Not a Number (NaN) sqrt(-42) = NaN 43

44 Floating Point Representation Double Precision Floating point: 64-bit representation: 1-bit sign 11-bit (biased) exponent 52-bit fraction (with implicit 1). double in Java, C, C++, S Exp Mantissa 1 11-bit 52 - bit 44

45 Ints and Floats int x = ; float f = x; System.out.println(x); x = f; System.out.println(x); Is this Java Legal? 45

46 Ints and Floats int x = ; float f = x; System.out.println(x); x = (int) f; System.out.println(x); Is this Java Legal? Java worries you might chop off a fraction Requires a cast Does not worry about f = x why not? 46

47 Ints and Floats int x = ; float f = x; System.out.println(x); x = (int) f; System.out.println(x); Program output: Why? 47

48 Ints and Floats int x = ; float f = x; System.out.println(x); x = (int) f; System.out.println(x); Program output: Hint x7314f x7314f900 Why? 48

49 Danger: floats cannot hold all ints! Many programmers think: Floats can represent all ints NOT true First summer internship I had: Need some floats and some ints: just use floats! Bug in their code! Other developers shocked as I demonstrated problem Doubles can represent all 32-bit ints (but not all 64-bit ints) S Exp Mantissa 1 11-bit 52 - bit Exp Mantissa 1 8-bit 23- bit 49 0

50 What about characters/strings? Many important things stored as strings Your name What is a string? How should we represent strings? 50

51 ASCII Character Representation Oct. Chr. 000 nul 001 soh 002 stx 003 etx 004 eot 005 enq 006 ack 007 bel 010 bs 011 ht 012 nl 013 vt 014 np 015 cr 016 so 017 si 020 dle 021 dc1 022 dc2 023 dc3 024 dc4 025 nak 026 syn 027 etb 030 can 031 em 032 sub 033 esc 034 fs 035 gs 036 rs 037 us 040 sp 041! 042 " 043 # 044 $ 045 % 046 & 047 ' 050 ( 051 ) 052 * , / : 073 ; 074 < 075 = 076 > 077? 101 A 102 B 103 C 104 D 105 E 106 F 107 G 110 H 111 I 112 J 113 K 114 L 115 M 116 N 117 O 120 P 121 Q 122 R 123 S 124 T 125 U 126 V 127 W 130 X 131 Y 132 Z 133 [ 134 \ 135 ] 136 ^ 137 _ 140 ` 141 a 142 b 143 c 144 d 145 e 146 f 147 g 150 h 151 i 152 j 153 k 154 l 155 m 156 n 157 o 160 p 161 q 162 r 163 s 164 t 165 u 166 v 167 w 170 x 171 y 172 z 173 { } 176 ~ 177 del Each character is represented by a 7-bit ASCII code. It is packed into 8-bits 51

52 Many types Unicode UTF-8: variable length encoding backward compatible with ASCII Linux UTF-16: variable length Windows, Java UTF-32: fixed length 52

53 Basic Data Types Bit: 0, 1 Bit String: sequence of bits of a particular length 4 bits is a nibble 8 bits is a byte 16 bits is a half-word 32 bits is a word 64 bits is a double-word Character: ASCII 7 bit code (packed in a byte) Decimal: (BCD code) digits 0-9 encoded as 0000 thru 1001 two decimal digits packed per 8 bit byte Integers: 2's Complement (32-bit representation). Floating Point: Single Precision (32-bit representation). Double Precision (64-bit representation). Extended Precision(128 -bit representation). How many +/- #'s? Where is decimal pt? How are +/- exponents represented? 53

54 Next time Memory Pointers Arrays Strings Bitwise operations Instruction Set Architecture Reading Start Chapter 2 54