Computer Systems: A Programmer s Perspec4ve. Have a tour of computer system at first... Chapter 1

Transcription

1 Computer Systems: A Programmer s Perspec4ve Have a tour of computer system at first... Chapter 1 1

2 Computer System Runs the sojware and manages the hardware RISC vs CISC ADDRESS BUS DATA BUS SOFTWARE HARDWARE LOAD/STORE } Opera4ng System ETC ADDRESSIBILITY BIG/LITTLE ENDIAN PIPELINING ALIGNMENT ISA 2

3 Outline Opera7ng System So;ware Hardware 3

4 The role of the opera4ng system Protect the computer from misuse Provide an abstrac7on for using the hardware so that programs can be wriben for a variety of different hardware Manage the resources to allow for reasonable use by all users and programs on a computer 4

5 The UNIX Opera4ng System Developed in 1970s at Bell Labs Kernel wriben in C, also developed at the same 7me C was developed for the purpose of wri7ng UNIX and systems programming We are using a variant of UNIX named Linux Other UNIX variants exist, such as Solaris, and the various BSDs (OpenBSD, NetBSD, FreeBSD, OSX) 5

6 Linux - OS hbps:// 6

7 Outline Opera7ng System So;ware Hardware 7

8 Text/Ascii A file is a sequence of bytes - not a magical container holding the bytes, but the bytes themselves How this informa7on is treated depends on the context the same sequence of bits can be used to represent a character, or an integer, or a floa7ng- point number, or an instruc7on, or... It's all a maber of interpreta7on % emacs hellot.c & 8

9 The compila4on system revisited hello.c %gcc - o hello hello.c Type in program using an editor of your choice (file.c); plain text.c +.h =.i which is the ul7mate source code? i.e. # includes expanded and #defines replaced.i à.s which is assembler source code.s à.o which is an object file; fragments of machine code with unresolved symbols i.e. some addresses not yet known (vars/subrs). hello.o + library links à a.out (default name); resolves symbols, generates an executable. %hello 9

10 Why assembly language? Instruc7on based execu7on Each program on a computer is a sequence of instruc7ons wriben in machine language Processor executes one instruc7on at a 7me in a program, then executes the next one in turn To study code in this form, it's helpful to use assembly language rather than machine language code gcc S hellot.c 10

11 Assembly language really?! Chances are, you ll never write programs in assembly Compilers are much beber & more pa7ent than you are But: Understanding assembly is key to machine- level execu7on model Behavior of programs in presence of bugs High- level language models break down Tuning program performance Understand op7miza7ons done/not- done by the compiler Understanding sources of program inefficiency Implemen7ng system so;ware Compiler has machine code as target Opera7ng systems must manage process state Crea7ng / figh7ng malware x86 assembly is the language of choice! 11

12 Another way to get assembly code Disassembler A tool that determines the instruc7on sequence represented by an executable program file Unix command gcc o hellot hellot.c objdump D t s hellot - d, - - disassemble Display assembler contents of executable sec7ons - D, - - disassemble- all Display assembler contents of all sec7ons - S, - - source Intermix source code with disassembly - s, - - full- contents Display the full contents of all sec7ons requested - t, - - syms Display the contents of the symbol table(s) - T, - - dynamic- syms Display the contents of the dynamic symbol table 12

13 Outline Opera7ng System So;ware Hardware 13

14 Hardware Organiza4on (big picture) Register File hellot executable stored on disk* PC ALU System bus Memory bus Expansion slots for other devices such as network adapters, video cards, etc. Bus Interface I/O bridge Main memory I/O bus USB controller Mouse Keyboard Graphics adapter Display Disk Controller Disk* 14

15 HW organiza4on details Processor (CPU) Interprets/executes instruc7ons stored in main memory Updates the PC to point to the next instruc7on PC (Program Counter) points at (contains the address of) some machine- language instruc7on in main memory ALU Computes new data and address values Register file Small storage device that consists of a collec7on of word- sized registers, each with their own name ISA instruc7on set architecture defines The processor state The format of the instruc7ons The effect each instruc7on will have on the state Instruc7ons: hbp:// jump.com/cis77/reference/instruc7ons_by_mnemonic.html 15

16 HW organiza4on details (cont.) I/O Devices System s connec7on to the external world Transfers informa7on back and forth between the I/O bus and an I/O device Main Memory Temporary storage Holds both the program and the data it manipulates Von Neumann architecture Is organized as a linear array of bytes each with its own unique address star7ng at zero Bus Transfers one word at a 7me Fundamental system parameter Amount can fetch from memory at one 7me Tends to be the size of the data bus 16

17 Memory Hierarchy (chp. 6) 9/27/17 17

18 Target of Memory Hierarchy Op4miza4ons Reduce memory latency The latency of a memory access is the 7me (usually in cycles) between a memory request and its comple7on Maximize memory bandwidth Bandwidth is the amount of useful data that can be retrieved over a 7me interval Manage overhead Cost of performing op7miza7on (e.g., copying) should be less than an7cipated gain 9/27/17 18

19 Abstrac4on Provided by the OS Process (chp. 8) The running of a program done by the processor Threads = mul7ple execu7on units Includes memory and I/O device (i.e. files abstrac7on) Virtual Memory (chp. 9) Provides each process with the illusion that is has exclusive use of the main memory Program code and data Includes files Begins at same fixed address for all processes Address space (chp. 7) Files (chp. 10) Sequence of bytes 19

20 Address Space a quick look An array of 8- bit bytes A pointer is just an index into this array ADDRESS SPACE Kernel virtual memory Decrip4on/info Memory invisible to user code User stack (created at run 7me) Implements func7on calls Memory mapped region for shared libraries Ex. priny func7on 32/64 bit star4ng address Run- 7me heap (created at run 7me by malloc/ calloc) Read/write data Read- only code and data } Dynamic in size Program (executable file) Fixed size Address 0 No7ce symbolically drawn with memory star7ng at the bobom 20

21 What is a system? A collec4on of intertwined hardware and systems sojware that must cooperate in order to achieve the ul4mate goal of running applica4on programs 21

22 Informa4on Representa4on and Interpreta4on Chapter 2 22

23 Outline General introduc7on Hexadecimal and other nota7ons Addressing and byte order Reading Assignment: Chapter 2 Sec7on

24 Are you sure? 8 Let s check * see sizeck.c * try m32 op4on ANSI rules Variables of type char are guaranteed to always be one byte. There is no maximum size for a type, but the following rela7onships must hold: sizeof(short) <= sizeof(int) <= sizeof(long) sizeof(float) <= sizeof(double) <= sizeof(long double) 24

25 Number of values (vs range of values) Every computer has a word size Nominal size of integer and pointer data Address space depends on word size à 2 word- size- in- #bits Is it big enough? 64- bit high- end machines becoming more prevalent Portability issues insensi7ve to sizes of different data types # bytes # bits # of values (2 #bits) low high E E E E E E E E E E E E+38 25

26 Interpreta4on & Representa4on What does this mean? abc x61 Representa7on (encode?) ASCII Simple Binary* One s complement Two s complement* Binary Coded Decimal Floa7ng- point* Limited number of bits to encode a value Will there be a 4me that the value we want to encode does not fit? Yes! OVERFLOW Need to be aware of the range of values that each limited number of bits will hold Inaccuracies exist (see overflw.c) 26

27 Floa4ng point Google what every computer scien7st should know about floa7ng point Squeezing infinitely many real numbers into a finite number of bits requires an approximate representa7on (rounding) Overflow à + Not associa7ve Due to finite precision of the representa7on A float has roughly seven decimal digits of precision see floatpt.c 27

28 Informa4on Storage (general) The machine level program generated has no informa7on about data types It s the C compiler that maintains this type of informa7on The different mathema7cal proper7es of integers vs floa7ng- point arithme7c stem from the difference in how they handle the finite- ness of their representa7ons Integers smaller values but more precise Real wide range of values, but only approximately Defines ranges of values Computer security vulnerabili7es Need to know/understand before progress to machine level programming (chp.3) 28

29 Informa4on Storage (details) Byte = smallest addressable unit of memory Virtual memory = very large array of bytes Address = how byte of memory is uniquely iden7fied Virtual address space = the set of all possible addresses Reminder: no data typing at this level 29

30 Outline General introduc7on Hexadecimal and other nota7ons Addressing and byte order 30

31 Hexadecimal Nota4on (Hex) Base 16 Useful in describing bit paberns Digits 0-9 and A- F In C 0x or 0X prefix interpreted as hex value Not case sensi7ve Example FA1D37B > 0Xfa1d37b, 0xFA1D37B, 0xfA1d37B Easy to convert to/from hex, octal and binary DEC HEX Notes ^ ^ ^ ^ A 11 B 12 C 13 D 14 E 15 F 31

32 Binary to Hex to Octal FYI: bit stands for binary digit Fact: 2 4 = 16 and 2 3 = 8 The power is the # of bits per hex/octal digit Binary to Hex Every 4 bits = 1 hex digit Octal base 8 Digits 0-7 Binary to Octal Every 3 bits = 1 octal digit Example: FA1D37B 16 DEC OCT HEX BIN Notes A B C D E F

33 When value x 10 is a power of 2 x = 2 n for some non- nega7ve integer n, then if x = = 2 4 = = Binary rep of x à 1 followed by n zeroes Hex rep of x is à n = i + 4j Reminder: hex digit 0 in binary is 0000 Leading hex digit where 0 <= i <=3 1 (i=0) 2 (i=1) 4 (i=2) 8 (i=3) Followed by j hex 0s Examples: n=9 à so i = 1 and j = 2 à x = (2)(00) hex i.e. x = n=6 à so i = 2 and j = 1 à x = (4)(0) hex i.e. x =

34 Convert to Decimal (why?) Base ten (decimal): digits 0-9 E.g., = Base eight (octal): digits 0-7 E.g., = Base 16 (hexadecimal): digits 0-9 and A- F. 13C 16 = Base 2 (binary): digits 0, = In general, radix r representa7ons use the first r chars in {0 9, A...Z} and have the form d n- 1 d n- 2 d 1 d 0. Summing d n- 1 r n- 1 + d n- 2 r n d 0 r 0 converts to base

35 Every base is base 10 EXPLANATION In general, 10 X = X = = = = = = = = =

36 Base Conversions Base Conversions Convert to base 10 by mul)plica)on of powers = ( ) 10 Convert from base 10 by repeated division = ( ) 8 Conver7ng base x to base y: convert base x to base 10 then convert base 10 to base y 36

37 More prac4ce Convert from base = ( ) 3 and check = ( ) 16 and check Another way from decimal to base n for n = 2 n 8 n 7 n 6 n 5 n 4 n 3 n 2 n 1 n From LEFT TO RIGHT, ask how many and subtract (219) 10 = ( ) 2 = ( ) 16 37

38 Hex and Binary addi4on/subtrac4on Hex add first, then convert hex to binary and add A + 8 = 13 + F = BEAD = Subtract in hex first, then convert each value to binary and subtract 5CD2 2A0 = = A8D2 3DAC = à carry/borrow 16 each 7me, since the next place is 16 7mes as large (see prac7ce problems) 38

39 Outline General introduc7on Hexadecimal and other nota7ons Addressing and byte order 39

40 Addressing and byte ordering Two conven7ons (for mul7- byte objects)* What is the address of the object? What is the order of the bytes in memory? Typically Mul7- byte objects are stored con7guously The address of the object is given as the smallest address of the bytes used Ø 4- byte integer stored as hex value at address 0x100 Ø So &x = 0x100, and Ø The 4 bytes of x would be stored at memory loca7ons 0x100, 0x101, 0x102, 0x103 xx xx xx xx * Does not apply to characters because they are single byte values 40

41 Addressing and byte ordering (cont) Big Endian The big end goes at byte zero big end means the most significant byte of the given value LiBle Endian The lible end goes at byte zero lible end means the least significant byte of the given value Byte zero means the smallest address used to store the given value Example: hex/given value is 0x What is the big end of the given value? à 01 What is the lible end of the given value? à 67 What is the lower memory address i.e. byte zero? 0x100 Byte order 0x100 0x101 0x102 0x103 Big Endian Litle Endian

42 Big/Litle Endian one lible two lible three lible endians ;o) Byte order Big Endian Litle Endian 0x x x x Byte order Big Endian Litle Endian 0x x x x There is no technological reason to choose one byte ordering conven7on over the other Need to choose a conven7on and be consistent Typically invisible to most applica7on programmers as results are iden7cal What if transferring data, though? Need to know when looking at integer data in memory 42

43 X86 is litle endian Largely, for the same reason you start at the least significant digit (the right end) when you add because carries propagate toward the more significant digits. Pu ng the least significant byte first allows the processor to get started on the add a;er having read only the first byte of an offset. A;er you've done enough assembly coding and debugging you may come to the conclusion that it's not lible endian that's the strange choice it's odd that we humans use big endian. A side note: Humans mostly read numbers and only some7mes use them for calcula7on. Furthermore we o;en don't need the exact numbers when dealing with large quan77es - taking that into account - big endian is a sensible choice for humans It reflects the difference between considering memory to always be organized a byte at a 7me versus considering it to be organized a unit at a 7me, where the size of the unit can vary (byte, word, dword, etc.) 43

44 Endian- ness When is byte ordering an issue? 1. Communica7ons over a network between different machines 2. Representa7on of integer/real numeric data 3. Circumven7on of normal type system Ø Using a cast to allow an object to be referenced according to a different data type from which it was created Use and even necessary for system- level programming Ø Can cast such that the value is a sequence of bytes rather than an object of the original data type #include <stdio.h> void main() { int x = 0x , i; unsigned char *xptr = &x; priny("the integer x is 0x%x\n",x); for (i = 0; i < 4; i++) priny("byte %d is %.2x\n",i+1, *(xptr+i)); } (castex.c) 44

45 Systems can have different word size byte sizes for each type endian- ness (for numeric values) representa7on of pointers character encoding schemes (ascii, ebcdic, unicode) instruc7on formats (again, just a sequence of bytes) ETC 45