Università degli Studi di Siena

HIGH PERFORMANCE COMPUTER ARCHITECTURE

AF monodisc.

Scheda dell'insegnamento

Anno accademico

2016/2017

Anno accademico di espletamento

2016/2017

Tipologia di insegnamento

Caratterizzante

Afferenza

Corso di Corso di Laurea Magistrale in COMPUTER AND AUTOMATION ENGINEERING – INGEGNERIA INFORMATICA E DELL’AUTOMAZIONE

Dipartimento Ingegneria dell’informazione e scienze matematiche

Settore disciplinare

SISTEMI DI ELABORAZIONE DELLE INFORMAZIONI (ING-INF/05)

Lingua

Inglese

Crediti

Anno di corso

Docenti

GIORGI ROBERTO

Ciclo

Primo Semestre

Ore di attivita' frontale

Ore di studio individuale

135

Prerequisiti

Computer Architecture.
Programming Fundamentals.

Obiettivi formativi

Knowledge of the techniques for supporting the parallelism in computer systems. Ability to implement parallel applications.

Contenuti

Introduction to parallel computing. Multicore and multithreaded processors. Scaling and Law of Amdhal. Branch Prediction. Speculative execution. Superscalar Architecture. Out-of-order execution. Introduction to VLIW architectures. Ordering of memory accesses. Multimedia/Vector Operations. Parallel programming. Programming for application-specific processors and graphics (GPU). Introduction to reconfigurable logic. Programming in distributed environments (OpenMP and MPI). Datacenters.

Programma esteso

* Introduction, Pipeline review.
* Dynamic Instruction Scheduling
* Tomasulo: An Efficient Algorithm for Exploiting Multiple Arithmetic Units
* Branch Prediction: speculation of branch condition and branch target.
* Predictor types, Bimodal, BHSR, BHT, PHT, 2-level adaptive, GAg, PAg, PAs. Other predictors (gshare, gselect, tournament).
* Introduction to Superscalar Processor: general scheme and Renaming.
* Full-System Simulator for single-core and multi-core systems.
* Superscalar execution example: Re-Order Buffer and Instruction Window. Case studies: MIPS, Alpha, AMD, Pentium.
* Software methods to extract Instruction Level Parallelism.
* Introduction to multiprocessor systems, Flynn's taxonomy, UMA, NUMA, COMA systems, programming models
* Coherence Protocols: Write Update, Write Invalidate, Hybrid. Snoopy based protocols: the MESI and DRAGON protocols
* Memory Consistency Models: Sequential Consistency and Relaxed Consistency
* Introduction to parallel programming.
* Parallelsim and Performance
* CUDA Architecture and Programming Elements
* Advanced topics: reconfigurable computing; datacenters.

Bibliografia consigliata

* MAIN TEXT: M. Dubois, M. Annavaram, P. Stenstrom, "Parallel Computer Organization and Design", Cambridge University Press, 2012, ISBN: 978-0-521-88675-8

* J.L. Hennessy, D.A. Patterson, "Computer Architecture: A Quantitative Approach" 5th Edition, Morgan Kaufman/Elsevier, 2012, ISBN: 978-0-12-383872-8.

* D. Culler, J.P. Singh, A. Gupta, "Parallel Computer Architecture: A Hardware/Software Approach", Morgan Kaufman/Elsevier, 1998, ISBN-10 1558603433.

* M.J. Flynn, "Computer Architecture: Pipelined and Parallel Processor Design", Jones and Bartlett Publishers, Inc., 1995, ISBN 0867202041

* David B. Kirk and Wen-mei W. Hwu, "Programming Massively Parallel Processors: A Hands-on Approach", Morgan Kaufmann (February 5, 2010) ISBN 0123814723.

Modalità di erogazione

Convenzionale

Metodi didattici

The course includes a theoritical part and a practical part (about 50%). A large part of the hours will be carried out in laboroatory to experience the possibility of estimating the performance through simulators.

Metodi di valutazione

Tipo di esame

Scritto e Orale Separati

Modalita' di verifica dell'apprendimento

Written examination followed by oral examination. Passing (mark non less then 18/30) the written mid_term+final_term permits to avoid the written exam for the first two regular exams of the academic year (the oral part must be performed in any case). Optionally, you can present a group project (max 3 people) which contributes to the final mark up to 10/30 to be added to the final mark obtained after written and oral exam, in case of a technical project. Instead of the project one can present a technical paper discussion which contributes up to 4/30.

Valutazione

Voto Finale

Calendario attività didattiche

Data di inizio del periodo didattico

Lunedì, 3 Ottobre, 2016

Data di fine del periodo didattico

Venerdì, 20 Gennaio, 2017

Altre informazioni

Both in the written and oral exams (but also in the projects) the student is mainly required to show his/her detailed understanding of the topic, at least at the level shown by the teacher during the lesson. It's greatly appreciated the capacity of reasoning on the problem, rather then a mechanical (pedant) description of the topic.
In case of written exercise, we mainly look at the correctness of the solution (in terms of numbers) and a very short justification of the chosen way to carry out the exercise (lengthy general wording is completely unnecessary).
In case of oral question, the topic is typically one of the concepts illustrated during the lesson. Elements that are required are, for instance: the proof of the concept/theorem, precise schematic of the system, detailed behavior and functioning, reasons why this solution is used in the real-world.

Altri insegnamenti

Anno di corso: 1

ADVANCED DATABASE SYSTEMS
6 crediti - Caratterizzante
ADVANCED DIGITAL IMAGE PROCESSING
9 crediti - Affine/Integrativa
AUTOMATA AND QUEUEING SYSTEMS
6 crediti - Caratterizzante
HIGH PERFORMANCE COMPUTER ARCHITECTURE
9 crediti - Caratterizzante
MODELS AND LANGUAGES FOR BIOINFORMATICS
6 crediti - Affine/Integrativa

Anno di corso: 2

DATA ANALYSIS
6 crediti - Caratterizzante
DESIGN OF APPLICATIONS, SERVICES AND SYSTEMS
9 crediti - Caratterizzante
HUMAN-CENTERED ROBOTICS
6 crediti - Caratterizzante
MULTIVARIABLE AND NONLINEAR CONTROL
6 crediti - Caratterizzante

Tu sei qui

HIGH PERFORMANCE COMPUTER ARCHITECTURE

HIGH PERFORMANCE COMPUTER ARCHITECTURE