Mechanical Engineering MS Project Presentation by Mr. Peter B. Khoury
When: Friday,
December 9, 2016
4:00 PM
-
5:00 PM
Where: Textiles Building 101E
Description: Mechanical Engineering MS
PROJECT PRESENTATION by Mr. Peter B. Khoury
DATE:
December 9, 2016
TIME:
4:00 p.m. – 5:00 p.m.
LOCATION:
Textile Building, Room 101E
TOPIC:
Investigating the Computational Efficiency of a Non-Hydrostatic Ocean Model on the Massachusetts Green High Performance Computing Center
ABSTRACT:
Scientists today use computer models to study oceanic variability. The majority of these models are still unable to capture all of the complex physical interactions that exist in the ocean. Improving the computational speed of these models remains a high priority. In this project, we use a three-dimensional non-hydrostatic ocean model which evolves momentum, temperature, and salinity in time. We explore the model performance for different grid sizes and compile-time options on the Massachusetts Green High Performance Computing Center. For smaller grid sizes we find serial computations (no OpenMP) faster than parallel computations with optimization level O3 for the Intel fortran compiler. Changing the optimization level to O2 gives superior performance for parallel compared to serial computations. For larger grid sizes however, there is no clear preference for serial or parallel. Amongst the serial runs, we find that increasing the number of threads from 4 to 10 improves the efficiency of the code regardless of the grid size. On the other hand, the parallel runs yield similar performance for both 4 and 10 threads. For smaller grids using 10 threads gave a better performance for parallel runs with optimization level O2 for the Intel fortran compiler. For larger grids, there was no trend with respect to the number of threads affecting the code efficiency. For the optimization level O3, both 4 and 10 threads resulted in almost similar performance regardless of the grid size.
ADVISOR:
Dr. Amit Tandon, Department of Mechanical Engineering
(atandon@umassd.edu, 508-999-8357)
COMMITTEE MEMBERS:
Dr. Mehdi Raessi, MNE Department of Mechanical Engineering
Dr. Sanjiv Ramachandran, MNE Department of Mechanical Engineering
Open to the public. All MNE students are encouraged to attend.
For more information, please contact Sue Cunha (scunha@umassd.edu, 508-999-8492).
PROJECT PRESENTATION by Mr. Peter B. Khoury
DATE:
December 9, 2016
TIME:
4:00 p.m. – 5:00 p.m.
LOCATION:
Textile Building, Room 101E
TOPIC:
Investigating the Computational Efficiency of a Non-Hydrostatic Ocean Model on the Massachusetts Green High Performance Computing Center
ABSTRACT:
Scientists today use computer models to study oceanic variability. The majority of these models are still unable to capture all of the complex physical interactions that exist in the ocean. Improving the computational speed of these models remains a high priority. In this project, we use a three-dimensional non-hydrostatic ocean model which evolves momentum, temperature, and salinity in time. We explore the model performance for different grid sizes and compile-time options on the Massachusetts Green High Performance Computing Center. For smaller grid sizes we find serial computations (no OpenMP) faster than parallel computations with optimization level O3 for the Intel fortran compiler. Changing the optimization level to O2 gives superior performance for parallel compared to serial computations. For larger grid sizes however, there is no clear preference for serial or parallel. Amongst the serial runs, we find that increasing the number of threads from 4 to 10 improves the efficiency of the code regardless of the grid size. On the other hand, the parallel runs yield similar performance for both 4 and 10 threads. For smaller grids using 10 threads gave a better performance for parallel runs with optimization level O2 for the Intel fortran compiler. For larger grids, there was no trend with respect to the number of threads affecting the code efficiency. For the optimization level O3, both 4 and 10 threads resulted in almost similar performance regardless of the grid size.
ADVISOR:
Dr. Amit Tandon, Department of Mechanical Engineering
(atandon@umassd.edu, 508-999-8357)
COMMITTEE MEMBERS:
Dr. Mehdi Raessi, MNE Department of Mechanical Engineering
Dr. Sanjiv Ramachandran, MNE Department of Mechanical Engineering
Open to the public. All MNE students are encouraged to attend.
For more information, please contact Sue Cunha (scunha@umassd.edu, 508-999-8492).
Topical Areas: Faculty, General Public, Staff and Administrators, Students, Students, Graduate, Students, Undergraduate, University Community, College of Engineering, Mechanical Engineering, Lectures and Seminars