SI2-SSE: An Interactive Parallelization Tool

University Of Texas At Austin, Austin TX

Investigators

Abstract

Many scientists, engineers, and students use the programming languages C, C++, and Fortran. Scientists who use these languages to conduct their experiments set up their programs to run in sequence -- what computer scientists call "serial processing." In order to make the best use of the state-of-the-art High Performance Computing (HPC) platforms to run more efficient experiments or to run very large simulations, the programs written in these languages need to be able to run simultaneously, what computer scientists call "in parallel" as distinct from "serial". Enabling the parallelization of the applications and software, however, remains a challenge. This project proposes deployment of an Interactive Parallelization Tool (IPT), which helps scientists, engineers, and students build on their customary use of C/C++/Fortran to create more efficient programs that can run on advanced HPC platforms. The IPT will be made available to the community through a web-portal for convenient code generation and testing on computational resources of the national cyberinfrastructure (CI). The team has developed a prototype IPT to parallelize multiple design patterns written in C/C++ at a high-level of abstraction. The proposed project will harden the IPT prototype into a production-quality tool and will extend support for the efficient parallelization of additional types of C/C++ applications. The team will also prototype the parallelization of Fortran applications. Carefully selected test cases will be provided with the IPT to engage and educate the users about the advantages and disadvantages of the various parallel programming strategies so that they can be appropriately applied in solving real-world problems. The IPT can be extended to support additional parallel programming paradigms, applications classes, and languages. It will be evaluated using code from multiple domains (e.g. biology, geosciences, civil engineering, and astronomy), and potential users will be engaged during the testing phase. This will all be served through a web-based portal, and will enable a low-risk guided approach for researchers, practitioners, and students to incorporate HPC in their work.