Large scale numerical simulations via reconfigurable computing hardware: An application using discrete element modeling of granular material
Award year: 2005-2006
This project will explore the use of reconfigurable hardware and computing systems for large scale numerical simulations in civil engineering. Over the last several years we have developed advanced simulation codes including a code, DBLOKS3D for simulating flow of granular particles. DBLOK3D is a discrete element code developed at UIUC that a) uses polyhedral particle representation, b) includes new particle contact detection algorithms and c) compiles on multiple platform including Windows and Linux. The code is used to model granular material movement such as earthmoving equipment interaction with soils and simulation of loading of granular material. The code can be used to simulate large scale particle movement including landslides, avalanches, and bulk material handling. DBLOKS3D is computationally expensive; for example a 1 sec real time analysis of an assemblage of 30,000 particles requires about 6-8 hrs on a PC (3.2 GHz). The code was ported to NCSA's Altix SGI machines and was parallelized using OpenMp Shared memory environment. However the speedup was limited to about a factor of 4 which is insufficient given that large scale simulation with millions of particles are envisioned.
We propose to take advantage of concepts of reconfigurable hardware (and hardware acceleration) and the experimental systems available at NCSA to explore the placement of specific parts of the code onto field programmable gate arrays (FPGA) to significantly increase the speed of the code. This work will be done in close collaboration with the Innovative systems Lab (ISL) at NCSA. Success of this effort will open the door for implementation of other large scale engineering analysis codes that the PI has developed in a reconfigurable computing hardware environment.