Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
39 views

CO322: DS & A Parallel Computing: Dhammika Elkaduwe

This document discusses parallel computing and algorithms that can be implemented in parallel. It provides examples of bubble sort, matrix multiplication, and odd-even sort algorithms. Bubble sort cannot be parallelized as each loop depends on the previous one. Matrix multiplication has a runtime of O(n^3) and can be parallelized by distributing the multiplication operations across CPU cores. Odd-even sort can also be parallelized by having different CPU cores process the odd and even elements simultaneously. The key takeaway is that parallel systems require algorithms that can break down problems into independent sub-problems.

Uploaded by

Mohamaad Sihatth
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
39 views

CO322: DS & A Parallel Computing: Dhammika Elkaduwe

This document discusses parallel computing and algorithms that can be implemented in parallel. It provides examples of bubble sort, matrix multiplication, and odd-even sort algorithms. Bubble sort cannot be parallelized as each loop depends on the previous one. Matrix multiplication has a runtime of O(n^3) and can be parallelized by distributing the multiplication operations across CPU cores. Odd-even sort can also be parallelized by having different CPU cores process the odd and even elements simultaneously. The key takeaway is that parallel systems require algorithms that can break down problems into independent sub-problems.

Uploaded by

Mohamaad Sihatth
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

CO322: DS & A

Parallel Computing

Dhammika Elkaduwe

Department of Computer Engineering


Faculty of Engineering

University of Peradeniya

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Why?

World is parallel!
Your phone has few cores in it!!
Can we make use of more than one processing element?

Depends on the algorithm


I For some algorithms parallel processes are useless
I For some others you can make use of them

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Why?

World is parallel!
Your phone has few cores in it!!
Can we make use of more than one processing element?

Depends on the algorithm


I For some algorithms parallel processes are useless
I For some others you can make use of them

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Example algorithm: Bubble sort Java code

public void bubble_sort(int [] data) {


int i,j;
for(i=0; i < data.length; i++) {
for(j = data.length-1; j > i; j--) {
if(data[j] < data[j-1]) {
int tmp = data[j];
data[j] = data[j-1];
data[j-1] = tmp;
}
}
}
}

Can you make this implementation parallel?


No. Every loop depends on the previous one(s)!

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Example algorithm: Bubble sort Java code

public void bubble_sort(int [] data) {


int i,j;
for(i=0; i < data.length; i++) {
for(j = data.length-1; j > i; j--) {
if(data[j] < data[j-1]) {
int tmp = data[j];
data[j] = data[j-1];
data[j-1] = tmp;
}
}
}
}

Can you make this implementation parallel?


No. Every loop depends on the previous one(s)!

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Example algorithm: Matrix multiplication

For two matrices Aam and Bmb , if C = A × B then


m
P
Cij = Aik ∗ Bkj
k=1
static int [][] multiply(int [][]A, int [][]B) {
/* implement me!
* What is/are the preconditions? */
}

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Example algorithm: Matrix multiplication Java code

static int [][] multiply(int [][]A, int [][]B) {

int ai = A.length, aj=A[0].length, bi=B.length,


bj=B[0].length;
int C[][] = new int [ai][bj];
int i, j, k, sum;
if(aj != bj) /* may be panic */ return C;
for(i=0; i < ai; i++) {
for(j=0; j < bj; j++) {
for(k=0, sum=0; k < aj; k++) sum += (A[i][k] *
B[k][j]);
C[i][j] = sum;
}/* for{i */ } /* for(j= */
return C;
}

What is the run-time complexity? (Aa×n , Bn×b )


Can you make it parallel? Why?
Dhammika Elkaduwe CO322: DS & A Parallel Computing
Example algorithm: Odd-Even sort

Basic idea:
I Similar to bubble sort
I Can be implemented in parallel systems (you tell me why :))
I Compare all odd ones with adjacent element, repeat for even
elements.
I Repeat until array is sorted.

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Example algorithm: Odd-Even sort Java code

static void oddEvenSort(int [] data) {


boolean sorted = false;
while(!sorted) {
sorted = true;
for(int i=0; i< data.length-1; i +=2) {
I What is the
if(data[i] > data[i+1]) { worst time
swap(data, i, i+1); complexity?
sorted = false; I What is the
} /* if */ } /* for */ best time
complexity?
for(int i=1; i< data.length-1; i +=2) {
if(data[i] > data[i+1]) { I Can use run
swap(data, i, i+1); this parallel?
sorted = false; Why?
} /* if */ } /* for */
}
}

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Exercises

1. Out of bubble, insert, selection, quick and merge sort which


ones can be implemented to run parallelly?
2. Can you implement matrix multiplication on parallel system?
2.1 if so, what is the run time complexity on a system with 2 CUP
cores for Aa×n , Bn×b ?
2.2 What if I have m > n number of CPUs? will it help?

Dhammika Elkaduwe CO322: DS & A Parallel Computing


Take home point

I Parallel systems depends on algorithms


I Some algorithms can make use of parallel cores while other
cannot
I Many cores does not make sense without good algorithms
I Increase the speed by overlapping executions

Dhammika Elkaduwe CO322: DS & A Parallel Computing

You might also like