Topics Covered After Midterm - PowerPoint PPT Presentation

1 / 7
About This Presentation
Title:

Topics Covered After Midterm

Description:

most questions will cover these topics, but will use also pre-midterm knowledge ... single source shortest paths discussion on Johnson's algorithm ... – PowerPoint PPT presentation

Number of Views:65
Avg rating:3.0/5.0
Slides: 8
Provided by: Pao3
Category:

less

Transcript and Presenter's Notes

Title: Topics Covered After Midterm


1
Topics Covered After Midterm
  • Parallel Sorting
  • Parallel Matrix Algorithms
  • Parallel Graph Algoritms
  • most questions will cover these topics, but will
    use also pre-midterm knowledge
  • example question
  • give a PRAM algorithm for back substitution

2
Sorting Summary
  • Sequential complexity Q(n log n) (comparison
    based)
  • Some useful concepts
  • Compare Split
  • 0-1 Sorting Lemma
  • Sorting networks
  • Sorting algorithms covered
  • Odd-even transposition sort (parallel bubble
    sort)
  • Shear sort (sorting algorithm for 2D mesh)
  • Bitonic sort (simple, relatively efficient
    sorting network)
  • Parallel Quick Sort (no comment)
  • Sample Sort (bucket sort variation)

3
Possible Questions
  • for which of the following algorithms does 0-1
    Sorting Lemma apply?
  • is it true that it applies to all sorting
    networks?
  • understand how each of the sorting algorithms
    works
  • so you can simulate it on a given input
  • you can write pseudocode for it
  • understand its strong points and limitations
  • perform bitonic split or merge on given
    sequences
  • how would you do bitonic sort on a ring network?
    What would be the cost?
  • identify the communication patterns used in a
    given sorting algorithm
  • given input sequence, what would be the global
    pivots in the sample sort?

4
Matrix Algorithms Summary
  • Matrix-Vector Multiplication
  • with 1D/2D partitioning
  • Matrix-Matrix Multiplication
  • simple parallel algorithm
  • Cannons algorithm
  • the DNS algorithm
  • Solving Systems of Linear Equations
  • simple Gaussian elimination
  • with pipelining/partial pivoting
  • back substitution

5
Matrix Algorithms Possible Questions
  • understand how each algorithm works
  • be able to simulate it
  • write pseudocode for processor pi (or pi,j)
  • determine communication patterns and costs
  • memory requirements/efficiency
  • load balancing issues

6
Graph Algorithms
  • Minimal Spanning Tree
  • Prims algorithm
  • Shortest Paths
  • single source Dijkstras algorithm
  • all- pair Dijkstras, Floyds algorithm
  • Transitive Closure
  • modified Floyds algorithm
  • Connected Components
  • merging partial forests
  • Algorithms for Sparse Graphs
  • maximal independent set Lubys algotihm
  • single source shortest paths discussion on
    Johnsons algorithm

7
Graph Algorithms Possible Questions
  • Understand the algorithms
  • be able to simulate them
  • write pseudocode
  • be able to apply the techniques to similar
    problems
Write a Comment
User Comments (0)
About PowerShow.com