Hybrid MPI-OpenMP versus MPI Implementations:  A Case Study

Osvaldo Mangual; Marvi Teixeira; Reynaldo Lopez-Roig; Felix Javier Nevarez-Ayala

Download Paper | Permalink

Conference: 2014 ASEE Annual Conference & Exposition
Location: Indianapolis, Indiana
Publication Date: June 15, 2014
Start Date: June 15, 2014
End Date: June 18, 2014
ISSN: 2153-5965
Conference Session: Curricular Issues in Computing and Information Technology Programs II
Tagged Division: Computing & Information Technology
Page Count: 18
Page Numbers: 24.680.1 - 24.680.18
DOI: 10.18260/1-2--20571
Permanent URL: https://peer.asee.org/20571
Download Count: 611

Paper Authors

biography

Osvaldo Mangual Honeywell Aerospace

visit author page

Osvaldo Mangual has a Bachelor degree in Computer Engineering and a Master degree in Electrical Engineering from Polytechnic University of Puerto Rico in the area of Digital Signal Processing. He currently works at Honeywell Aerospace, PR, as a FPGA and ASIC Designer.

visit author page

biography

Marvi Teixeira Polytechnic University of Puerto Rico

visit author page

Dr. Teixeira is a Professor at the Electrical and Computer Engineering and Computer Science Department at Polytechnic University of Puerto Rico
He holds a Ph.D. and MSEE degrees from University of Puerto Rico at Mayaguez and a BSEE degree from Polytechnic University.
Professor Teixeira is an IEEE Senior Member, a Registered Professional Engineer and a former ASEE-Navy Summer Faculty Fellow.

visit author page

biography

Reynaldo Lopez-Roig Polytechnic University of Puerto Rico

visit author page

Mr. Lopez received his B.S. in Computer Engineering from the Polytechnic University of Puerto Rico in 2013. His work as an undergraduate research assistant was related to the implementation and benchmarking of parallel signal processing algorithms in clusters and multicore architectures. Mr. Lopez is currently working at the Jet Propulsion Laboratory as a Software Systems Engineer.

visit author page

biography

Felix Javier Nevarez-Ayala Polytechnic University of Puerto Rico

visit author page

Felix Nevarez is an Associate Professor in the Electrical and Computer Engineering Department at the Polytechnic University of Puerto Rico, He earned a BS in Physics and a MS in Electrical Engineering from the University of Puerto Rico at Mayaguez and a BS in Electrical Engineering from the Polytechnic University of Puerto Rico. HIs research interests are in numerical simulations and parallel and distributed programming.

visit author page

Download Paper | Permalink

Abstract

Hybrid MPI-OpenMP versus MPI Implementations: A Case Study We explored the performance of a hybrid MPI-OpenMP parallel implementation versus adirect MPI implementation in a 64-processor cluster architecture featuring 16 nodes with 4 coresper node. A scalability study was carried out where we varied the signal length for two differentsets of parallel cores. The algorithm being benchmarked is a parallel cyclic convolution algorithm with nointerprocessor communication that tightly matches our particular cluster architecture. In thisparticular case study a time-domain-based cyclic convolution algorithm was used for eachparallel subsection. By using MPI for distributing the data to the nodes, and then using OpenMPfor distributing the data among the cores inside each node we can match the architecture of ouralgorithm to the architecture of the cluster. Each core processes and identical program withdifferent data using a single program multiple data (SPMD) approach. All pre and post-processing tasks were performed at the master node. We first partitioned the cyclic convolution into 16 parallel subsections of one-fourth theoriginal convolution length using a radix-4 approach. This implementation was tackled usingMPI where the data was distributed from the master node to 16 parallel cores. We then repeatedthe execution by using an MPI-OpenMP approach. Using this hybrid technique data of one-halfthe original length was distributed from the master node to four nodes using MPI while the finalexecution was performed under OpenMP at each of the four nodes by halving each subsection intwo and using all four processors in the multicore node. The final processing length is again one-fourth of the original signal length and both method use 16 processors. We repeated the processfor eight different signal lengths. We then used the algorithm to partition the cyclic convolution into 64 parallelsubsections of one-eighth the original convolution length using a radix-8 approach. Thisimplementation was tackled using MPI where the data was distributed from the master node to64 parallel cores. We then repeated the execution by using an MPI-OpenMP approach. Usingthis hybrid technique data of one-fourth the original length was distributed to 16 nodes usingMPI while the final execution was performed under OpenMP at each of the 16 nodes by halvingeach subsection in two and using the four processors in the multicore node. The final processinglength is again one-eighth of the original signal length and both method use 64 processors. Werepeated the process for the same eight different signal lengths plus two additional lengthsafforded by the greater memory available when using 64 cores. We found that the MPI implementation had a slightly better performance than the hybrid,MPI-OpenMP implementation. We established that the speedup increases very slowly, in favorof the MPI-only approach, as the signal size increases. This is consistent with what is reported inthe literature. As a future work we plan to further our code optimization efforts and to benchmark formemory efficiency, where the hybrid approach could have an advantage, as well as for increasedperformance.

Citation
Format

Mangual, O., & Teixeira, M., & Lopez-Roig, R., & Nevarez-Ayala, F. J. (2014, June), Hybrid MPI-OpenMP versus MPI Implementations: A Case Study Paper presented at 2014 ASEE Annual Conference & Exposition, Indianapolis, Indiana. 10.18260/1-2--20571

TY - CPAPER
AB - Hybrid MPI-OpenMP versus MPI Implementations: A Case Study We explored the performance of a hybrid MPI-OpenMP parallel implementation versus adirect MPI implementation in a 64-processor cluster architecture featuring 16 nodes with 4 coresper node. A scalability study was carried out where we varied the signal length for two differentsets of parallel cores. The algorithm being benchmarked is a parallel cyclic convolution algorithm with nointerprocessor communication that tightly matches our particular cluster architecture. In thisparticular case study a time-domain-based cyclic convolution algorithm was used for eachparallel subsection. By using MPI for distributing the data to the nodes, and then using OpenMPfor distributing the data among the cores inside each node we can match the architecture of ouralgorithm to the architecture of the cluster. Each core processes and identical program withdifferent data using a single program multiple data (SPMD) approach. All pre and post-processing tasks were performed at the master node. We first partitioned the cyclic convolution into 16 parallel subsections of one-fourth theoriginal convolution length using a radix-4 approach. This implementation was tackled usingMPI where the data was distributed from the master node to 16 parallel cores. We then repeatedthe execution by using an MPI-OpenMP approach. Using this hybrid technique data of one-halfthe original length was distributed from the master node to four nodes using MPI while the finalexecution was performed under OpenMP at each of the four nodes by halving each subsection intwo and using all four processors in the multicore node. The final processing length is again one-fourth of the original signal length and both method use 16 processors. We repeated the processfor eight different signal lengths. We then used the algorithm to partition the cyclic convolution into 64 parallelsubsections of one-eighth the original convolution length using a radix-8 approach. Thisimplementation was tackled using MPI where the data was distributed from the master node to64 parallel cores. We then repeated the execution by using an MPI-OpenMP approach. Usingthis hybrid technique data of one-fourth the original length was distributed to 16 nodes usingMPI while the final execution was performed under OpenMP at each of the 16 nodes by halvingeach subsection in two and using the four processors in the multicore node. The final processinglength is again one-eighth of the original signal length and both method use 64 processors. Werepeated the process for the same eight different signal lengths plus two additional lengthsafforded by the greater memory available when using 64 cores. We found that the MPI implementation had a slightly better performance than the hybrid,MPI-OpenMP implementation. We established that the speedup increases very slowly, in favorof the MPI-only approach, as the signal size increases. This is consistent with what is reported inthe literature. As a future work we plan to further our code optimization efforts and to benchmark formemory efficiency, where the hybrid approach could have an advantage, as well as for increasedperformance.
AU - Osvaldo Mangual
AU - Marvi Teixeira
AU - Reynaldo Lopez-Roig
AU - Felix Javier Nevarez-Ayala
CY - Indianapolis, Indiana
DA - 2014/06/15
PB - ASEE Conferences
TI - Hybrid MPI-OpenMP versus MPI Implementations: A Case Study
UR - https://peer.asee.org/20571
DO - 10.18260/1-2--20571
ER -