Description

UNIVERSIDAD AUTÓNOMA DE MADRID

MASTER’S DEGREE IN RESEARCH AND INNOVATION IN COMPUTATIONAL INTELLIGENCE AND INTERACTIVE SYSTEMS (I2-ICSI)

Laboratory 3: LOOP PARALLELIZATION WITH OPENMP

Compulsory assignment (7 points)

1. Download associated material (openmp3.tar.gz) from Moodle’s course page into personal working directory.
2. Uncompress and untar associated material:

gunzip openmp3.tar.gz tar –xvf openmp3.tar

3. Go to directory openmp3/task1
4. Edit and understand example “task1.c”. The sequential loop (function Loop) is the one parallelized in unit 2.4.6.
5. Using OpenMP, implement the parallel version of the above sequential loop (function LoopPar) based on the pseudocode derived and shown in unit 2.4.6. Be aware that OpenMP parallel loops have implicit barrier synchronization at the end, thus making explicit barriers redundant. Also notice that outer parallel loops in nested loops must explicitly declare as private the loop variables of their corresponding inner loops.
6. Compile the program and execute it, making sure that both loops yield the same result (program must print out “Equal 1”).
8. Write down the results and conclusions to lab report, including the source code of function LoopPar (homework).
1

9. Go to directory openmp3/task2
10. Edit and understand example “task2.c”. The sequential loop (function Loop) is the one listed below:
for (i=0; i<N; i++) for (j=0; j<N; j++)
{
B[i+1][j-1] = A[i+1][j] – A[i-1][j]; // S1
D[i][j] = D[i-1][j-1] * 3; // S2
C[i+1][j] = D[i][j] * B[i-1][j] * 2; // S3
C[i-1][j+1] = D[i+2][j+1] * 3; // S4
}
11. Implement the parallel version of the above sequential loop using OpenMP (function LoopPar).
12. Compile the program and execute it, making sure that both loops yield the same result (program must print out “Equal 1”).
13. Write down the sequential and parallel wall times for two and four threads, respectively. Compute the associated speedups and efficiencies. Check if the parallel program yields superlinear speedup.
14. Write down the results and conclusions to lab report, including the dependency graph and the source code of function LoopPar (homework).

Optional assignment (3 points)
15. Go to directory openmp3/task3
16. Edit and understand example “task3.c”. The sequential loop (function Loop) is the one listed below:
for (i=0; i<N; i++) for (j=0; j<N; j++)
{
C[i+1][j+1] = D[i-3][j+1] * 3; // S1
D[i-1][j] = D[i][j] * B[i][j-1] – 2; // S2
B[i+1][j-1] = A[i+1][j] – A[i-1][j]; // S3
A[i][j] = D[i-1][j-1] * 3; // S4
}
17. Implement the parallel version of the above sequential loop using OpenMP (function LoopPar).
18. Compile the program and execute it, making sure that both loops yield the same result (program must print out “Equal 1”).
19. Write down the sequential and parallel wall times for two and four threads, respectively. Compute the associated speedups and efficiencies. Check if the parallel program yields superlinear speedup.
20. Write down the results and conclusions to lab report, including the dependency graph and the source code of function LoopPar (homework).

2