Scribd
Upload
21 views2 pages

Lab 3

The document discusses parallelizing the odd-even sorting algorithm using OpenMP. It describes the sequential odd-even sorting algorithm and how to parallelize it in two ways, by grouping comparisons across threads or assigning each comparison to a thread. The tasks involve implementing sequential and parallel versions, recording times for different problem sizes and thread counts, and calculating speedups to analyze efficiency.

Uploaded by

ims
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
21 views2 pages

Lab 3

The document discusses parallelizing the odd-even sorting algorithm using OpenMP. It describes the sequential odd-even sorting algorithm and how to parallelize it in two ways, by grouping comparisons across threads or assigning each comparison to a thread. The tasks involve implementing sequential and parallel versions, recording times for different problem sizes and thread counts, and calculating speedups to analyze efficiency.

Uploaded by

ims
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Lab 3.

OpenMP – parallel sorting and performance studies

Main goals of the assignment


• Learn about the differences between the efficiency and the improvement percentage using parallel version.

• Learn how to use a low number of processing elements in simulation of a large number of parallel tasks

The problem to solve


General overview
A parallel version of odd-even sorting that is efficient is requested.

Background
The odd-even sorting compares every 2 consecutive numbers in the array and swap them if first is greater than the second
to get an ascending order array. It consists of 2 phases – the odd phase and even phase:
Odd phase: Every odd indexed element is compared with the next even indexed element.

Even phase: Every even indexed element is compared with the next odd indexed element.
See the bellow figure for an example:

Figure 1: Example

How to parallelize
The maximum degree of parallelism is n/2. Usually the number of processing elements is lower than n/2. Then we need
to try an almost equal distribution of the coomparisons-exchanges on our cores. How we can do?

1
To do
1. Write the sequential code to implement odd-even sorting.
2. Write the OpenMP versions:

(a) group the n/2 comparisons-exchanges of one parallel step on the p threads;
(b) create a thread for each comparison-exchange (do not take into account the fact that we have p  n by using
n/2 the number of threads)
Introduce time records (hint: omp get wtime) before and after the part that is parallelized. Decide which version is
optimal using a very large sequence to be sorted (milions order).
(n)
3. Record the times Tp for the best variant in table like the folowing (with the maximum cores that you have, e.g. 8
or 16) – see the table

Table 1: Put inside the boxes the recorded times


n\p 1 2 4 8
100000
200000
400000

4. In order to compute the speedup we need to know the sorting time for the best sequential algorithm (quicksort!)
Compute the speedups using
(n)(quick−sort)
T
Sp(n)(odd−even) = 1(n)(odd−even)
Tp
and record them in a similar table with the above one.
5. Draw conclusions related to:

• is the parallel version of odd-even effcient?


• dependence of the answer on the problem dimension n.

You might also like