Abstract

This thesis proposes a solution to streamline the process of using supercomputing re- sources on Southern Methodist University’s ManeFrame II supercomputer. A large segment of the research community that uses ManeFrame II belong outside of the computer science department and the Lyle School of Engineering. While users know how to apply compu- tation to their field, their knowledge does not necessarily extend to the suite of tools and operating system that are required to use ManeFrame II. To solve this, the thesis proposes an interface for those who have little knowledge of Linux and SLURM to be able to use the supercomputing resources that SMU’s Center for Scientific Computation provides.

OpenMP is a compiler extension for C, C++ and Fortran that generates a binary using multithreading using in-code directives. With knowledge of OpenMP, researchers are already able to split their code into multiple threads of execution. However, because of the complexity of Linux and SLURM, using OpenMP with the supercomputer can be problematic. This thesis focuses on the user of ANTLR, a programming language recognition tool. This tool allows for the insertion of directives into code which serves to generate batch files that are compatible with the supercomputer scheduling software, SLURM. With the batch file, the user is then able to submit their code to the supercomputer.

Additional tools around this core piece of software facilitate a usable interface. In order to make the tool accessible to those without a background in software, the proposed forward- facing solution is a user interface to upload their code and returns a batch file that the user can use to run their code. This eliminates the need for a new user to download, compile and run the ANTLR distribution to generate a batch file.

By abstracting away these complexities into a web interface, the solution can generate a batch submission file for the user. Additional tooling assists the user in finding empty nodes for code execution, testing the compilation of their code on the supercomputer and running a timed sample of their code to ensure that OpenMP is leading to a speedup in execution time.

Degree Date

Winter 12-15-2018

Document Type

Thesis

Degree Name

M.S.

Department

Computer Science and Engineering

Advisor

Francis Coyle

Subject Area

Computer Science

Number of Pages

54

Format

.pdf

Creative Commons License

Creative Commons Attribution-Noncommercial 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License

Download

Included in

Engineering Commons

Share

COinS