Exam 1 Info - Open Computing Facilityrfguo/CS61CL/CS61CL Guest... · 2009-03-06 · Exam 1 Info Exam 1 returned in Tu/W lab cs61cl-tb@h30 [4] ~ > glookup -s exam1 Number of grades

Exam 1 Info

Exam 1 returned in Tu/W lab

cs61cl-tb@h30 [4] ~ > glookup -s exam1

Number of grades reported: 169Mean: 43.6Standard deviation: 10.8Minimum: 2.01st quartile: 37.02nd quartile (median): 46.03rd quartile: 52.0Maximum: 59.0Max possible: 60.0

Reminder: Homework 6 is due Monday, March 9 at 11:59pm Project 2 has been posted (due Friday, March 20) Exam solutions posted on next Monday (probably)

Images Used with Permission from Wikipedia

CS61CL Spring 2009 EditionRichard Guo

Guest Lecture - CALL

Where are we going? Number Representation Floating Point Compilation, Assembly,

Linking, Loading Caches



Motivation

What happens when we type gcc program.c o program?

What work is done to turn the source code into something the computer can process?

How is it possible to play N64 games on your PC?



Lowdown



Compiler

Translates from one programming language to another (e.g. C > Assembly)

Pseudoinstructions may be present in output Targeted optimization at this step



Assembler

Decodes assembly language into machine language (opcodes + symbol table)

Splits pseudoinstructions into actual ones Resolves symbolic names Processes directives Generates symbol and relocation tables Output is in an object file



Assembler

Regular Instructions

Pseudoinstructions

Symbolic Names

Relocation Table

Symbol Table

Directives

Straightup conversion Instruction: sra $s1 $0 8

R Fields: 0 0 0 17 8 3 Binary: 000000 00000 00000

10001 01000 000011 Hex: 0x00008A03

What about la $a0, str?



Assembler


Pseudoinstructions

Symbolic Names

Relocation Table

Symbol Table

Directives

la $a0, str

Pseudo, so broken down into: lui $at,left_16(str)ori $a0,$at,right_16(str)

Note the usage of $at



Assembler


Pseudoinstructions

Symbolic Names

Relocation Table

Symbol Table

Directives

L1: slt $t0, $0, $a1 beq $t0, $0, L2 addi $a1, $a1, -1 j L1 L2: add $t1, $a0, $a1

We can resolve branches right now! L1: slt $t0, $0, $a1 beq $t0, $0, 2 addi $a1, $a1, -1 j L1 L2: add $t1, $a0, $a1



Assembler


Pseudoinstructions

Symbolic Names

Relocation Table

Symbol Table

Directives

Jumps are more troublesome Require the absolute address Don't know how objects will arrange Forward addressing

Thus, maintain two tables Symbol Table Relocation Table



Assembler


Pseudoinstructions

Symbolic Names

Relocation Table

Symbol Table

Directives

Relocation table contains a list of things we need to fix in the file Labels referenced in jump instructions Global labels targeted by la External labels not in the current file

We fix these later on when the information becomes available



Assembler


Pseudoinstructions

Symbolic Names

Relocation Table

Symbol Table

Directives

Symbol table associates identifiers with where it is declared Relative address of labels to the start

of the text segment Ordering of variables in the .data

segment

This is for later reference by the program or by another object file



Assembler


Pseudoinstructions

Symbolic Names

Relocation Table

Symbol Table

Directives

Directives give guidelines to assembler

.globl – Declares a global variable

.text – Marks beginning of the code segment

.data – Start of variable declaration for storage in memory

.asciiz – Declares a \0 terminated string

.word – 32 bit integer

And many more...



ELF Layout

The Executable and Linking Format Object Header Section Headers .text segment .data segment Relocation Table Symbol Table Misc. + Debug Info



Linker



Linker

Input has all the necessary files Start by concatenating data and text segments Then fix up things in all the relocation tables by

consulting the symbol table of each file Assume 0 is the start address of text segment Separate object files allow

Distribution of obfuscated object+header files Quick recompilation of just the modified source files



Loader

Loads text and data of program into memory Initialize registers ($sp, $gp, $fp, etc.) Moves command line input parameters to registers Executes and increments program counter



Summary

Compiler turns source code into assembly code (.c > .s)

Assembler turns assembly code into machine code (.s > .o)

Linker combines many object files and libraries into an executable (.o + .o > a.out)

Loader loads the program into memory and runs (./a.out)



Demo

Questions? We'll clarify here with a demo



Compiler vs Interpreter

Compiler translates highlevel language into machine language (eventually)

Examples: Assembly, C/C++, Java Interpreter runs the highlevel language directly Examples: Matlab, Perl, Python, Scheme, Java

Compilation and interpretation are not exclusive Java translates source to bytecode, then interprets



Compiler vs Interpreter

(+) Easier to write interpreter (CS 61A), less code (+) Interpreter is more easily portable () Interpreter has overhead

Must maintain internally symbols, stack, and state

() Programs on interpreters are harder to distribute



Just In Time Compilation

Compile only what we need when we encounter it Slow at first, fast later Combines the (+) of compilation and interpretation Allows runtime optimization for specific processor

instructions Java and Microsoft .NET do this



Dynamic vs Static Linking

So far, we've done static linking Embeds libraries and objects in a single file

Dynamic linking loads required objects at runtime (+) Reduced file size and memory usage at runtime (+) Library updates are propagated automatically () Linking process takes time

Documents

Exam 1 Info - Open Computing Facilityrfguo/CS61CL/CS61CL Guest... · 2009-03-06 · Exam 1 Info Exam 1 returned in Tu/W lab cs61cl-tb@h30 [4] ~ > glookup -s exam1 Number of grades