Implementing Parallel Graph Algorithms Spring 2015 Implementing Parallel Graph Algorithms Lecture 1: Introduction Roman Manevich Ben-Gurion University

Agenda Course overview Go technical Get back to explaining requirements in detail 2

Course Goals Learn basic techniques for parallelizing graph algorithms – Focus on the Galois system Experience by implementing one algorithm – Experimenting with different optimizations – Gathering and analyzing experimental data – Self-learning Write a short research paper-like report Present a short talk in final class Teams of 1-2 students 3

Grade 10% quality of final report 10% quality of presentation 5% code quality and code documentation 5% thoroughness of experimental evaluation 70% quality of sequential and parallel implementation – Demonstrate knowledge of algorithm – Experiment with different schedulers and different code structures – Memory allocation – Well-chosen data structures – Efficient synchronization – Result validation 4

Schedule TaskMeeting / Date 1.Introduction to parallel graph algorithms programming and Galois Today: Meeting 1 2.Get acquainted with Galois and modern C++ programming 3.Run single-source shortest path (SSSP) on road network input graph 4.Connect to server nx360srv1-3 4.Algorithm proposal15/4: Meeting 2 5.Implement sequential version of algorithm with Galois 6.Implement parallel version 7.Submit final report 8.Present results in a short (10 min.) presentation24/6: Meeting 3 5

My role Basic introduction to theory and practice Most of the work is by self learning but I can help with your individual projects Office hours: Wednesday 12:00-14:00 Course web-pageweb-page – Announcements – Forum 6

How to succeed in this course Start immediately – You will encounter many technical problems and stand no chance to solve them if you postpone Make sure you understand your algorithm well before starting to implement Make sure you understand why your optimizations maintain correctness Don’t get stuck for a very long time – ask me or schedule a meeting Be ethical Joe (a day before assignment deadline): “I don’t really understand what you want from me in this assignment, can you help me/extend the deadline”? 7

Algorithms Social networks – PageRank – Community detection Recommender systems – Collaborative filtering Programming languages – Dominators – Interprocedural analysis Computational geometry Generic – Graph coloring 8

Parallel server nx360srv1-3 (Linux) CPUs: 20 Intel Xeon cores of 2.8GHz (hyperthreaded can run 40 threads) RAM: 128GB Work remotely Lab staff will establish time splitting later 9

Next Introduction to basic parallelism and ADP (Amorphous Data Parallelism) Technical introduction to Galois 10