SC is the International Conference for
 High Performnance Computing, Networking, Storage and Analysis

SCHEDULE: NOV 13-19, 2010

Evaluation of Two Parallel Finite Element Implementations of the Time-Dependent Advection Diffusion Problem: CUDA versus MPI

SESSION: Research Poster Reception


TIME: 5:15PM - 7:00PM

AUTHOR(S):Lucas Veronese, Leonardo Lima, Alberto De Souza, Lucia Catabriga

ROOM:Main Lobby

We examine two parallel finite element implementations of the 2D time-dependent advection diffusion problem, one for multi-core clusters and one for CUDA enabled GPUs, and compare their performances. We used the SUPG stabilized finite element formulation to solve the corresponding 2D time-dependent advection diffusion equation. To address the resulting system of ordinary differential equations towards steady-state, we employed the implicit predictor/multicorrector scheme. The sparse linear system of each time-step (stored in a Compressed Storage Row - CSR - scheme in both implementations) was solved by the GMRES method. In our multi-core cluster implementation we used the MPI library for inter-core communication. Both implementations were written in C -- the CUDA version was derived from the multi-core version. Our experimental results showed that a desktop machine with a few CUDA enabled GPUs can achieve higher performance than a 24-machine (96 cores) cluster in this class of finite element problems.

Chair/Author Details:

Lucas Veronese - Universidade Federal do Espírito Santo

Leonardo Lima - Universidade Federal do Espírito Santo

Alberto De Souza - Universidade Federal do Espírito Santo

Lucia Catabriga - Universidade Federal do Espírito Santo

Add to iCal  Click here to download .ics calendar file

Add to Outlook  Click here to download .vcs calendar file

Add to Google Calendarss  Click here to add event to your Google Calendar

   Sponsors    IEEE    ACM