Profiling and Evaluation of Implicit and Explicit Storm Surge Models

Abstract

Abstract-Storm surge models are vital nowadays because the recent destruction caused by hurricanes. In this study, we are profiling and evaluating explicit and implicit solvers, two popular methods used in storm surge predictions. Hurricane Katrina storm surge hindcast has been used in this simulation for both methods. We use analysis tools to examine both methods from various perspectives including: execution times, computational intensity, characterization of instructions, cache memory misses and others. Many experiments have been performed with various number of processors. We did not only present the differences between methods, but also we addressed the challenges associated with these methods.Index Terms-Storm surge, implicit and explicit solvers, parallelization, and profiling.

Implementation of an Implicit Solver in ADCIRC Storm Surge Model

“…A total of 128 processors are used in all experiments below unless stated otherwise. Note that there are slight differences from the findings with previous papers [5,20]; these are attributed to the major update that RENCI environment has undergone recently. All variables are set to be the same for all three solvers to assure that the three solvers are tested under the same conditions, with the exception of solver-specific variables.…”