TDF: A compact file format plugin for FEniCS

Abstract

The Finite Element Method (FEM) is an important numerical method to solve Partial Differential Equations (PDEs). The widely used open source FEM platform — FEniCS, supports Hierarchical Data Format 5 (HDF5), which is very effective at storing and organizing large amounts of simulation data. However, HDF5 files can become prohibitively large for certain engineering applications, such as the thermal simulation of CPUs. Thus, this paper introduces a Type-Length-Value data format (TDF) plugin for compact storage of the FEM simulation solutions. Our Type-Length-Value (TLV) encoding implementation can be readily expanded to store more generic mathematical data generated by engineering and scientific applications. The evaluation results indicate that the TDF format could lead to a ≈95% space savings in our illustrated example. Meanwhile, the read/write speed has been improved compared to HDF5 used by FEniCS.

Keywords

• Hierarchical Data Format 5 (HDF5)
• Type-Length-Value (TLV)
• Finite Element Method (FEM)
• FEniCS