Computer Architecture and Organization Course at Princeton University, Study Guides, Projects, Research of Design and Analysis of Algorithms

Download Computer Architecture and Organization Course at Princeton University and more Study Guides, Projects, Research Design and Analysis of Algorithms in PDF only on Docsity! 1 Lecture 1: Introduction COS 375 / ELE 375 Computer Architecture and Organization Princeton University Fall 2015 Prof. David August 2 The Usual Suspects Me: Prof. David August, 221 CS Building, august@ Office Hours: M/W after class and by appointment TAs: Bochao Wang, E-Quad C-319B, bochaow@ Office Hours: Th/F 10:30-11:30AM Debajit Bhattacharya, E-Quad C-319D, dbhattac@ Office Hours: T/Th 3-4PM Hansen Zhang, 241 CS Building, hansenz@ Office Hours: M/W 3-4PM Course Objectives •  Enable you to design and build a computer •  Get a lump of matter to do your bidding •  Understand and evaluate tradeoffs in design •  Appreciate theory vs. practice •  Become a better implementer of algorithms •  Learn assembly/machine language programming •  Essential for OS and Compiler Work •  Essential for understanding processor design •  Understand modern Computer Organization •  High-level languages -> execution on physical material •  Help you to revolutionize computing •  Discuss the latest research •  Contribute some of our own? 4 1. Performance Evaluation •  Measures of performance •  Benchmarks and metrics 2. Instruction Set Architecture •  Instruction formats & semantics •  Addressing modes 3. Machine Arithmetic •  ALU design •  Integer multiplication & division •  Floating-point arithmetic 4. Processor Design •  Datapath design •  Instruction exec. & sequencing •  Hardwired & microcode control •  Pipelining 5. Hardware Design Languages •  Design with a Verilog •  Modeling and simulation 6. Memory Hierarchy •  Cache design & evaluation •  Virtual addressing •  Performance evaluation 7. Input/Output •  Types of I/O devices •  Device access and interface •  Device control •  I/O performance 8. Multiprocessor •  Interconnection networks •  Programming issues Course Topics 5 The Course Project •  Groups of 4 students – Sign up in September •  Work can be done anywhere •  Project consists of two parts •  ARM-based Instructions Set Simulator (in C) •  Implementation of processor onto an FPGA (in Verilog) 6 Pick a number 1,2,3 If the random number is the picture of a processor, then we have a quiz. 7 Quizzes/Homework Quizzes •  Chance quiz at the beginning of one class each week •  Not intended as a scare tactic – liberally graded •  Helps assess progress of class •  Just one question usually Homework •  4 homework sets •  Questions resemble exam questions 8 Exams •  Exams cover concepts presented in the lecture material, homework assignments, and/or required readings •  Sheet of paper allowed Midterm Exam •  Wednesday before Fall Break •  In class Final Exam •  The final exam will be cumulative, three hours in length •  Time/Place determined by the Registrar 19 Why Computer Architecture? 20 It’s a Wednesday: Pick a number 1,2,3 21 Quiz 0: Background (use index cards) Front: 1.  Full name and Email Address above the red line 2.  Major/UG or G/Year (immediately below the red line) 3.  Area (G: Research Area/UG: Interests) 4.  Briefly describe any C/C++ experience. 5.  In which programming languages are you fluent? 6.  What is a bit? Back: 1.  What is an instruction cache? 2.  What is the difference between a sequential and a combinational circuit? 3.  What is a MUX? 4.  Using AND, OR, and NOT gates design an XOR gate. 22 23 What is a Computer? Computers haven
t changed in over 50 years! •  Universal Turing Machine Equivalence •  Given enough time/memory, nothing new Computers have undergone enormous changes! •  New applications enabled •  New form factors Computers process information •  Input/Output •  State •  Computation Dude, your get’n a… EDSAC Universal Turing Machine “Thin. Light. Epic.” 24 Modern Computer Organization •  Where does the hard disk go? •  Computer system design •  Enable applications (speed, reliability, efficiency) •  Reduce cost (die size, technology, time-to-market) •  The Key: Manage Complexity! Processor Computer Control Datapath Memory Devices Input Output keyboard, mouse printer, speakers, monitor Connections for Information flow Coordination for proper operation 25 Abstraction •  Separate implementation from specification •  INTERFACE: specify the provided services. •  IMPLEMENTATION: provide code or HW for operations. •  CLIENT: code or HW that uses services. •  Examples: ADTs •  Principle of least privilege The Living Daylights: Bond and Saunders are in a house waiting for General Koskov to defect. Bond is preparing to shoot a sniper. Bond: What's your escape route? Saunders: Sorry old man. Section 26 paragraph 5, that information is on a need-to- know basis only. I'm sure you'll understand. 26 Intuition Client Interface - universal remote - volume - change channel - adjust picture - decode NTSC, PAL signals Implementation - cathode ray tube - electron gun - Sony Wega 36XBR250 - 241 pounds, $2,699 27 Intuition Client Interface - universal remote - volume - change channel - adjust picture - decode NTSC, PAL signals Implementation - gas plasma monitor - Pioneer PDP-502MX - wall mountable - 4 inches deep - $19,995 Can substitute better implementation without changing client! Interfaces in Computer Systems Instruction Set Architecture Applications Operating System FirmwareCompiler Instruction Set Processor I/O System Datapath & Control Digital Design Circuit Design Layout Software Hardware Hello World The Hello World Algorithm: 1.  Emit Hello World 2.  Terminate Java Program Hello World The Hello World Algorithm: 1.  Emit Hello World 2.  Terminate C Program Interfaces in Computer Systems Instruction Set Architecture Applications Operating System FirmwareCompiler Instruction Set Processor I/O System Datapath & Control Digital Design Circuit Design Layout Software Hardware 40 Physical Principles of Information • Fredkin-Toffoli axioms. E. F. Fredkin and T. Toffoli. Conservative logic. International Journal of Theoretical Physics, 21(3/4):219--253, 1982. •  The speed of propagation of information is bounded. • Speed of light • No action at a distance – causal effects propagate thru local interactions •  The amount of information which can be encoded in the state of a finite system is bounded. • Bounded by thermodynamical/quantum-mechanical considerations •  It is possible to construct macroscopic, dissipative physical devices which perform in a recognizable and reliable way the logical functions AND, OR, and FAN-OUT. Physical Devices Control In Out As we will see later, choice of device is critical!! Layout/Circuit Design 4-bit Adder 43 Digital Design 4-bit Adder 44 Digital Design 4-bit Adder Digital Design 4-bit Adder x3 x2 x1 x0 y3 y2 y1 y0 z3 z2 z1 z0 + Datapath 1 PC + Memory W Data R Data Addr W IR t s d op Adder Datapath is a conduit for information flow through the processor Complete Datapath > 0 = 0 1 Memory W Data R Data Addr W 2 Registers W Data W Addr A Addr B Addr A Data B Data W 80 PC + Cond Eval IR t s d op pc for jal store data addr pc + 1 pc for branch, jump result of arithmetic, logic, or addr for load addr addr for loads, stores 3 A L U load 8 0 16 8 16 8 Adder 48 Control (from the back of a napkin) The Hardware/Software Interface Instruction Set Architecture Applications Operating System FirmwareCompiler Instruction Set Processor I/O System Datapath & Control Digital Design Circuit Design Layout Software Hardware The Instruction Set Architecture The vocabulary of commands •  Defined by the Architecture (x86) •  Implemented by the Machine (Pentium 4, 3.06 GHz) •  An Abstraction Layer: The Hardware/Software Interface •  Architecture has longevity over implementation •  Example: add r1 = r2 + r3 (assembly) 001 001 010 011 (binary) Opcode (verb) Operands (nouns) 59 Slice of History – Intel Processors Pentium III •  9,5000,000 transistors •  125 mm2 •  450 MHz •  Introduced in 1999 60 Slice of History – Intel Processors Core 2 Duo (Merom) •  293,000,000 transistors •  143 mm2 •  1.6 GHz - 3.16 GHz •  Introduced in 2006 61 Slice of History – Intel Processors "Grove giveth and Gates taketh away." - Bob Metcalfe (inventor of Ethernet) Other Technology Trends •  Processor •  Logic Capacity: ~30% increase per year •  Clock Rate: ~20% increase per year •  Memory •  DRAM Capacity: ~60% increase per year •  Memory Speed: ~10% increase per year •  Cost per Bit: ~25% decrease per year •  Disk •  Capacity: ~60% increase per year Trends… 64 Trends? Is computer architecture dead? One More Trend… •  Intel 4004 (Thousands of transistors) •  Number of Designers: 2 •  Intel Core i7 (Billions of transistors) •  Number of Designers: ~1500 •  How does Intel manage so many designers on one project? •  Architects •  Microarchitects •  Circuit designers •  Validation •  Software •  This trend is exponential. What does this mean? Summary •  Read Chapter 1 in H&P •  Abstraction in HW and SW to manage complexity •  ISA defines the Hardware Software Interface •  Technology influences implementations… Poor choice of device technology: The Vacuum Tube Supercomputer Centre — - Access from freeway Private rail yard CPU cooling towers Bios Building Central Processing Unit Control Building Bus Building Power supply - 6 steam turbines @ 18 GVA each Cooling pond/ coal delivery Network 1/0 Oil storage farm Interface Building Control Building #2 Buildings http:/Avww.ominous-valve.com/vtsc.html