760 likes | 896 Views
CS5365. Introduction. Syllabus. There is course website that contains the syllabus www.cs.ttu.edu/~andersen/cs5365. Learning Objects. 1. Understanding of different classes of computer architectures from the design perspective,
E N D
CS5365 Introduction
Syllabus • There is course website that contains the syllabus • www.cs.ttu.edu/~andersen/cs5365
Learning Objects • 1. Understanding of different classes of computer architectures from the design perspective, • 2. Familiarity with design issues in pipelined vector oriented machines, • 3. Understanding of cache schemes for multiprocessor systems, • 4. Familiarity with parallelization techniques that exploit specific architecture designs, such processor arrays, symmetric, and NUMA multiprocessor systems, • 5. Understanding of the analytical approach to the design and analysis of interconnection networks, • 6. Ability to address computer architecture research problems
Measuring Outcomes • Students will be given topics to present in class these topics will be directly related to the lecture material covered in class but will focus on either current research in the field of multiprocessor systems or state-of-the-art technologies in the field of multiprocessor systems. Each student will present 3 times, each presentation will require them to demonstrate an understanding of their topic of choice, rubric will be posted on the grading criteria used. In addition each student will submit a paper that assesses an assigned research paper.
Measuring Outcomes • Vendor manuals, web sites and research papers will be the primary source of information. • Each student presentation will be 30 minutes, there will be two presentations during one class session.
Ombudsman and Disability • http://www.depts.ttu.edu/ombudsman/
Ombudsman and Disability • Students with disabilities: Any student who, because of a disability, may require some special arrangements in order to meet course requirements should contact the instructor at per.andersen@ttu.edu, 742-3527 or Engineering Center, Room 306D, as soon as possible to request necessary accommodations. • Students should present appropriate verification from the Disabled Student Support Program (DSSP) in the Dean of Students Office. No requirement exists that accommodations be made prior to completion of this approved University process.
Academic Conduct • The College of Engineering fully subscribes to the Code of Student Conduct as published in the Texas Tech University's "Student Affairs Handbook."
Assignments • The purpose of the assignments is to provide historical background how theory was put into practice as well as educating students on the state of the art.
Assignments • A course rubic for grading will be posted on the course website • All presentation must be submitted for grading to the course drop box • Each student will be doing 3 presentation each worth 100 points of the final grade • There will be one scientific paper 3-6 pages IEEE format worth 60 points of the final grade
Assignments • After every presentation session and new series of topics will be posted • Students can submit there own topics for presentation. • Students not presenting will be expected to pose questions to the presenter. A record will be kept of questions and who asked them and this will be class participation.
Note • No exams, no tests, no final exam.
Why High Performance Computers? • What are some of the technologies that are in use or being developed?
Introduction • We are going to study architecture • Single processor systems as a review • Models of multiprocessor systems • Pipelining • Vector processors • Array processors • Multiprocessor systems
Introduction • If someone said to you “we have a problem that needs a computational system”, how would you choose a system? • What would be your thoughts?
Introduction • What kind of system can actually do the job? • Can the workstation on my desk do it might be one question you ask yourself.
Introduction • The text mentions performance and cost as two factors. • What are the goals related to performance and cost?
Introduction • Besides performance and cost what are some other factors?
Introduction • Space • Ease of programming • Power consumption • Weight • Size • Reliability • Availability • Scalability, upgradeability?
Performance • How might we measure performance?
Performance • We might quote the clock rate. • “I have a 3.0 Ghz machine” • What does this mean?
Instruction count • Ic = no. of instr./program = size of the program
Cycles/Instruction • CPI = Number of cycles required per instruction = cycles/instruction • So how many cycles does an instruction take?
Execution Time • Lets assume we have a CPI what is the execution time of a program?
Ic • How many instruction would there be in an program?
Ic • A good compiler will reduce the number of instructions via optimization
CPI • What about instructions per cycle? • CISC versus RISC
τ • Time/cycle what controls that factor?
τ • Chip designer
MIPS • Millions of Instructions are sometimes quoted. • MIPS = (instr./cycle)(cycles/sec) x 10-6 • Is this meaning full?
MFLOPS • What about floating point operations/second? • Is this meaningful?
Introduction • What is Moore’s Law?
Moore’s Law • Moore's law is the empirical observation that the complexity of integrated circuits, with respect to minimum component cost, doubles every 24 months. It is attributed to Gordon E. Moore, a co-founder of Intel.
Introduction • Four Generations of Computers • Vacuum tubes • Transistors • Medium-scale integration • Large-scale integration
Definitions • Computer Organization — functional components are selected and interconnected to form a computer system • Computer Architecture — deals with the structure and behavior of the computer system as seen by the user • Performance specifications for various components are developed. • Component interconnections are defined. • Instruction set, format and addressing modes are defined.
Definitions • Computer Design — components specifications are refined. System is implemented using hardware (CPU, memory, hardwired control), software (operating systems, language design, etc.,) and firmware elements (control units, special purpose units, etc.)
Other important components: • Communications • organization of components, interconnection networks, protocols, interface design, etc. • Algorithms • to solve tasks eciently and make optimal use of the system components • Performance • A high performance systems involves all: hardware, software, communications (bandwidth), and algorithms.
Introduction • Common computing models
Serial • Examples in every day life?
Pipelining • Examples in every day life?
Parallel • Examples in every day life?
Serial • If four subtasks are done to assemble an automobile, each taking time T to complete, what is the total time to assemble one automobile by a single server? • What about N automobiles?
Serial • What about four subtasks on a single processor system?
Serial • 4T for one automobile • 4NT for N automobiles • 4T for 4 subtasks on a single processors system plus overhead. • What kinds of overhead am I talking about?