Discipline: Computer Sciences and Information Management
Subcategory: Computer Science & Information Systems
Janica Gordon - Southern University and A&M College
Co-Author(s): Daniel Osei-Kuffuor, Lawrence Livermore National Lab; Jeffrey Hittinger, Lawrence Livermore National Lab
Assume that there is a technique that would accelerate a linear solver and reduce the energy consumption by cutting the overall computation time. The performance of 32-bit operations is often at least twice as fast as the performance of 64-bit operations. By using a combination of 32-bit and 64-bit floating point arithmetic, the performance of many linear algebra algorithms can be significantly enhanced while maintaining the 64-bit accuracy of the resulting solution. A code was designed and implemented for the mixed precision iterative refinement algorithm using the LU factorization with the use of the single precision and double precision. Single precision, double precision, and mixed precision iterative refinement were compared and analyzed.
Not SubmittedFunder Acknowledgement(s): Mr. Daniel Osei-Kuffuor, Lawrence Livermore National Laboratory Jeffrey Kittinger, Lawrence Livermore National Laboratory CCMS, Lawrence Livermore National Laboratory Southern University and A&M College
Faculty Advisor: Daniel Osei-Kuffuor, oseikuffuor1@llnl.gov
Role: I did all of the research in the project with the help of my mentor.