Intro To Computer Architecture

Computer Organization Intro to Computer Architecture

So what are we really learning? • How low-level software works • How hardware works – CPU in particular

• Where these 2 meet and work with each other

main: la $a0,str li $v0,4 Syscall li $v0, 10 Syscall

Why is this in any way necessary? • Primary reasons: – make you a better computer scientist – make you a better software engineer

• Additional reason: – make you a more interesting person in nerd circles

• Yes, people who don’t know this stuff are boring

$$$$$ Our Daily Reminder $$$$$ • You’re paying me to be here • This material cannot be learned in one night, or even in one week – please give me your brain

• This is your opportunity to learn this stuff – as Computer Scientists, you’ll have many opportunities to use it

$$$$$$$$$$$$$$$$$$$$$$$$$$$$$

Learn Something Today • How many of these things can you define? – – – – – – – – –

CPU Register Cache Instruction Set ISA Assembly Language Assembler MIPS MARS

What is a CPU? • Central Processing Unit • The “brain” of the computer • What does it do? – Fetches, Decodes, AND Executes instructions

• What are Instructions? – commands to a computer

The Cell CPU Up Close

By the way • These days there’s a whole CPU Cooling market

A CPU has multiple regions • Control Unit – decodes instructions and relays orders to rest of CPU

• Arithmetic Logic Unit – performs basic mathematical operations

• CPU Memory – Registers – Cache

• We’ll see how these work in Part 3 of the semester

A CPU has memory? • Yes, a little • Far less than RAM • Let’s see who remembers the last slide: – What are the 2 types of CPU memory? • Registers • Cache

What are Registers? • Memory for executing binary instructions • All variables used to execute an instruction must be loaded into Registers • Accessing/Changing data in Registers is very fast – Why? • because it is very close to control unit & ALU

• CPUs have very small Register memory – MIPS uses 32 Registers

How much memory does a Register store? • One word • What’s a word? – well, first let’s understand what bits & bytes are

What are bits? • Stands for binary digit • Radix – 2, all representations in 1s and 0s • Everything in computers is stored in binary (programs, data (number, string), pictures, …. • Bits are easily represented in hardware • An open switch means a 0, closed switch means 1 (or vice versa) • Switches can be electrical, mechanical, semiconductor, magnetic

What are bytes? • A group of 8 bits • Ex: 01010101 • Memory can usually be addressable via bytes – What does that mean? • we can get a byte of data from memory if we know the memory address • we can change a byte of data in memory if we know the memory address

So what’s a word? • A group of bytes • Word size is dependent on architecture – 32-bit machine vs. 64-bit machine – Ex: MIPS uses 32 bit words (4 bytes)

• The memory of the computer will deliver 1 word at a time – So? • the wider the word, the greater the processing power

– Why? • more data can be processed each step

• CPU Instructions are word sized

So how much memory is that? • For MIPS Registers: – 32 registers X – 32 bits per register _______________ – 1024 bits (256 bytes)

• Hmm, that doesn’t seem like much

What is Cache? • A memory block in the CPU • It duplicates part of memory in RAM – still much smaller than RAM – Ex: Intel Core 2 has 8 MB of cache

• Why does it do this? – cache can be accessed faster than RAM

• How does it do this? – when a program executes, parts are loaded into cache – when memory is addressed, cache is checked first • if found it’s called a cache “hit” • for cache misses, parts of cache might need to be replaced

Registers vs. Cache • Registers: – is very small – can be directly accessed by developer code • we can load data into them • we can get data from them • we can specify which registers to use (they have names)

• Cache – is slightly bigger – is used managed automatically by the CPU – developers don’t directly instruct the CPU to use cache memory

Memory Comparisons faster

CPU Registers CPU Cache RAM (main memory) Hard Disk

more space

What is an instruction set? • Part of a system’s Architecture • Instructions: the words used to command a computer • Instruction set: a computer’s vocabulary • Different computers may use different instruction sets • NOTE: CPUs are hard-wired with: – an instruction set – an implementation of each type of available instruction

Are all instruction sets the same? • No • There are many similarities among them • But they are all different • Code used by a CPU with one architecture is generally unusable by a CPU

Common Operations • What types of things must all Instruction Sets be able to do? – – – – –

load data from memory addresses write data to memory addresses add/subtract/multiply/divide numbers conditional branching etc.

• Each will have different ways of doing these

What is an ISA? • Instruction Set Architecture • Includes: – – – – –

instruction set registers arrangement memory addressing numeric formats etc.

• Determines how instructions are given to a computer on the lowest level

Some ISAs • ARM – cell phones • Cell – PlayStation 3 • IA-32 – x86 PCs • MIPS – embedded systems • PowerPC – XBOX 360, GameCube, Wii • SPARC – Sun SPARC Workstations • And many, many more of course

What is an assembly language? • An implementation of an instruction set • Assembly code is written in an assembly language • It gets translated into machine code (binary) – by assembler

• The machine code is executed by the machine

Who writes assembly code? • Compilers • Crazy people – who have lots of time on their hands – who register for classes like CSE 220 • Smart people – working on a project where the speed or size of the program is critically important – like game programmers – who register for classes like CSE 220

HL vs. Assembly Language Programs

Critically Important Code? • NOTE: a skilled Assembly Programmer can in many cases write code that is more efficient than what a compiler may create • Compilers have become smarter & smarter though • Compilers may have optimization options

Some C/C++ programmers will mix in some assembly float arg1, arg2, add ; printf( "Enter two numbers : " ); scanf( "%f%f", &arg1, &arg2 ); /* Perform floating point Addition */ __asm__("fld %1;“ "fld %2;“ "fadd;“ "fstp %0;":"=g"(add):"g"(arg1),"g"(arg2)); printf( "%f + %f = %f\n", arg1, arg2, add );

Ref: http://www.codeproject.com/KB/cpp/edujini_inline_asm.aspx

MIPS - Course Assembly Language • MIPS – “Microprocessor without Interlocked Pipeline Stages” – ISA used primarily in embedded systems • PlayStation 1 & 2, Nintendo64, & PSP • digital cameras, DVD players, routers, wireless phones, HDTVs, etc.

• Why MIPS? – has a long educational legacy – good textbook – still widely used • MIPS Architecture and Assembly Language Overview

MIPS • Designed in early 1980s by John Hennessy – our textbook author and Stony Brook CS alumni

• Is a Reduced Instruction Set Computer architecture (RISC) – as opposed to Complex Instruction Set Computer architecture (CISC) • once used by most PCs • starting with the 486 and Pentium-I processors Intel implemented a RISC core in its processors • more on RISC vs. CISC later this semester

Classes of computers • Desktop Computers • Servers • Supercomputers • Embedded Systems

What are embedded systems? • “Computers designed to run one application or one set of applications which is normally integrated with the hardware and delivered as a single system” • Good to know: – more of these than all the others combined – very wide range of price & performance – most users don’t even know they’re using a computer

• Any examples?

MARS – Course IDE • MARS – MIPS Assembler & Runtime Simulator – freely downloadable 2.1 MB JAR file – http://courses.missouristate.edu/KenVollmar/MARS/

• Online tutorial: – http://courses.missouristate.edu/KenVollmar/MARS/tutorial.htm

• All MIPS programming assignments should be done using MARS

Why don’t CPUs just use C?

Abstraction

What’s so good about abstraction? • CPU makers can make CPUs better – faster – more efficient – easier to program

• Same for high level language programmers – you write a program in C – a compiler translates the C code into MIPS – an assembler translates the MIPS code into bytecode

hello.c /* hello.c * Purpose - print hello */ #include <stdio.h> int main(void) { printf("Hello, world!\n"); return 0; }

hello.asm ## hello.asm - print out "hello world" # text Segment # .text # tells assembler program code starts here .globl main # defines label for execution start main: # execution starts here la $a0,str # put string address into a0 li $v0,4 # system call to print syscall # out a string li $v0, 10 # Load exit syscall value syscall # Exit # data segment # .data # tells assembler data segment begins here str: .asciiz "hello world\n" # declaration of a string

bytecode executed by hardware 00111100000000010001000000000001 00110100001001000000000000000000 00100000000000100000000000000100 00000000000000000000000000001100 00100000000000100000000000001010 00000000000000000000000000001100

What did you learn today? • How many of these things can you define? – – – – – – – – –

CPU Register Cache Instruction Set ISA Assembly Language Assembler MIPS MARS

Next Time • Basic MIPS program format • MIPS instructions • MIPS registers