Special Courses 1 (High Performance Computer Systems)

CSM211: HIGH PERFORMANCE COMPUTER CONTROL AND PROCESSING

Lecturer Oleg S. Kozlov, Associate Prof, Ph.D..

This course will look at parallel architecture from point of view of High Performance Computer (H&C) processing, and evaluate their usefulness accordingly. The course is devoted to studying how HPC get faster and better solution, and covers the principles and design techniques of parallel architectures. It also covers both theoretical and pragmatic issues related to parallel computer architecture. Students study main features and scailability of parallel computer systems, classification and evaluation of HPC. Also this course covers some theoretical issues of parallel computer machines and relate that theory to real HPC architectures.

1. Basic Concepts of Parallel Computing

2. Decomposing the Potentially Parallel

3. Message-passing concepts

4. Data Parallel Concepts

5. Flynn's Taxonomy

6. Theoretical Models of Parallel Computation

7. Pipelines

8. Memory Organization

9. Network Topology

10. High Performance Architectures

11. Performance Model for Vectors and Parallel Machines

12. Program Parallelism

13. HPC Software Standards

14. Evaluation of Parallel Performance

1. Hwang K. Advanced Computer Architecture. McGraw-Hill, 1993.

2. Stone H.S. High Performance Computer Architecture. Addison-Wesley, Reading, MA, 1993.

CSM212: CONCURRENT COMPUTING (PARALLEL PROGRAMMING)

Lecturer: Yuri STatarinov, Associate Prof., Ph.D..

The problem of machine resources insufficiency and development of parallel and distributed information processing. Parallelism as a fundamental characteristic of calculations. The process concepts. Metamodels of processes and models interpretations. Formal models of parallel programming. Specification means of parallel calculating processes. Asynchronous processes control in multiprogrammed environments. Asynchronous parallel processes interaction mechanisms. -Interaction of the processes using common resources and shared data. Classic problems of asynchronous parallel processes interaction. Structural organization and processes control in multiprogrammed environments. System means of asynchronous processes control in the UNIX environment. Parallel calculations and desequention of algorithms. Parallel data processing in the systems using mass parallelism. Parallel programming in the PARIX operational system environment.

1. Formal models of parallel programming

2. Computing processes equivalence. Processors attributes conception

3. Programmes parallel schemes: data searching scheme and control scheme

4. Keller's operator systems

5. Desequention problem. Setting objectives

6. Realization with queues

7. Processes interaction organization

8. Low level synchronous primitives. Typical tasks on asynchronous processes interaction

9. High level synchronous primitives

10. Methods of asynchronous processes interaction mechanisms specification

11. System means of processes control in the UNIX environment

12. Parallel information development in systems using mass parallelism

13. Parallel programming in the PARIX environment

Course Text

1. Karp R., Miller R. Parallel program shemata// Journal of Computer and System Sciences, vol.3, No.4, 1969.

2. Baer J. A survey of some theoretical aspects of multiprocessing// ACM Computer Surveys, vol.5, No.1,1973.

3. Parallel processing Systems// Edited byD.Evans. - Cambridge: Cambridge University Press, 1982.

4. Hromkovic J. Communication complexity and parallel computing// Texts in Theoretical Computer Science: At EATCS Series, 1997.

CSM213: COMPUTER SYSTEMS PERFORMANCE ANALYSIS

Lecturer Vladimir A. Kirjanchikov, Associate Prof., Ph.D..

This course provides basic modelling, simulation and analysis background to systems analysts s^ that they will understand performance metrics and terminology, properly design performan^ experiments, use different types of simulation and analytical models, such as Markov chains Retry nets and queuing models and correctly interpret results. Also it highlights experimenta design techniques specifically intended to reduce analysis costs.

1. Introduction and overview. Basic definitions.

2. Measurement techniques. Benchmark programs. Measurement tools. Trace generation, Measurement perturbations.

3. Analyzing measured data. Measures of central tendency and variability. Confidence intervals.

4. System modelling and performance prediction. Finite Markov chains. Petry nets. Special-purpose models. Hierarchical performance evaluation models.

5. Simulation. Types of simulations. Random number generation and testing* Generating nonuniform distributions.

6. Workload modelling. Natural and artificial models. Executive and nonexecutive models.

7. Bottleneck and performance tuning.

1. Raj Jain, The Art of Computer Systems Performance Analysis, John Wiley and Sons, Inc, 1991.

2. K. Kant, Introduction to Computer System Performance Evaluation, McGraw-Hill, Inc., New York, 1992.

CSM214: ADVANCED COMPUTER ARCHITECTURE

Lecturer: Audrey V. Anissimov, Associated prof., Ph.D.

Discusses current computer practices for hardware and basic microcomputer instruction sets assembler and I/O programming, interrupts and interfaces. 3 credit. Prerequisite: Programming on Pascal. Programming on N. Computer Science. The Discrete Structures.

1. Introduction. Concept about architecture

2. The block diagram of the uniprocessor computer

3. Hierarchy of memory. Speed and productivity

4. Cache memory.

5. Processors.

6. Personal computers.

7. Formats of instruction and system of instruction of processors

8. The block diagrams of advanced processors architectures

9. Principles of fulfilment of interruptions.

10. Interface modules of the computer.

11. Comparison of peculiarities of independent work of the computer and its work in structure of a network.

12. Classification of networks. The basic components of networks.

1. The Me Graw-Hill Computer Hand Book. Editor in Chief Harry Helms, McGraw-Hili, Inc., New York, 1983.