TY - GEN
T1 - The role of floating point precision in two- and three-dimensional high Rayleigh Bénard convection modeled on Fermi GPU
AU - Sanchez, David A.
AU - Yuen, David A.
AU - Sun, Yujun
AU - Wright, Grady B.
PY - 2011
Y1 - 2011
N2 - We have implemented a second-order finite difference method for two-dimensional and three-dimensional Rayleigh-Béanard thermal convection, corresponding to convection in the Earth's mantle, on a single Fermi GPU. These codes are written in C for CUDA, making heavy use of CUBLAS routines for efficiency, and achieve performance on the order of 535 GFLOP/s and 100 GFLOP/s in single-precision and 230 GLFOP/s and 70 GFLOP/s in double-precision. We explore the sensitivity of this model to word length, finding that global characteristics remain constant despite a change in precision. Specifically, we compare the divergence between single- and double-precision runs with exactly identical initial conditions to the divergence between double-precision runs whose initial conditions have been perturbed by Gaussian noise. Our finding is that large-scale quantitative behavior (Nusselt number, number of plumes, etc) does not vary among these samples. This observation suggests a saving in time and computing resources could be enjoyed by implementing certain problems in single-precision. This is also valuable to scientists using iterative methods, as convergence may be completely unaffected by change of precision before the last few iterations. A particular interest is developed in the context of young Earth mantle convection, where higher Rayleigh numbers require both a finer computational mesh and a shorter timestep to properly resolve dynamic, small-scale features - compounding time wasted by inefficient or overly conservative computational implementations.
AB - We have implemented a second-order finite difference method for two-dimensional and three-dimensional Rayleigh-Béanard thermal convection, corresponding to convection in the Earth's mantle, on a single Fermi GPU. These codes are written in C for CUDA, making heavy use of CUBLAS routines for efficiency, and achieve performance on the order of 535 GFLOP/s and 100 GFLOP/s in single-precision and 230 GLFOP/s and 70 GFLOP/s in double-precision. We explore the sensitivity of this model to word length, finding that global characteristics remain constant despite a change in precision. Specifically, we compare the divergence between single- and double-precision runs with exactly identical initial conditions to the divergence between double-precision runs whose initial conditions have been perturbed by Gaussian noise. Our finding is that large-scale quantitative behavior (Nusselt number, number of plumes, etc) does not vary among these samples. This observation suggests a saving in time and computing resources could be enjoyed by implementing certain problems in single-precision. This is also valuable to scientists using iterative methods, as convergence may be completely unaffected by change of precision before the last few iterations. A particular interest is developed in the context of young Earth mantle convection, where higher Rayleigh numbers require both a finer computational mesh and a shorter timestep to properly resolve dynamic, small-scale features - compounding time wasted by inefficient or overly conservative computational implementations.
UR - http://www.scopus.com/inward/record.url?scp=81455131899&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=81455131899&partnerID=8YFLogxK
U2 - 10.1109/CSE.2011.122
DO - 10.1109/CSE.2011.122
M3 - Conference contribution
AN - SCOPUS:81455131899
SN - 9780769544779
T3 - Proc. - 14th IEEE Int. Conf. on Computational Science and Engineering, CSE 2011 and 11th Int. Symp. on Pervasive Systems, Algorithms, and Networks, I-SPA 2011 and 10th IEEE Int. Conf. on IUCC 2011
SP - 606
EP - 610
BT - Proc. - 14th IEEE Int. Conf. on Computational Science and Engineering, CSE 2011 and 11th Int. Symp.on Pervasive Systems, Algorithms, and Networks, I-SPAN 2011 and 10th IEEE Int. Conf. IUCC 2011
T2 - 14th IEEE Int. Conf. on Computational Science and Engineering, CSE 2011, the 11th International Symposium on Pervasive Systems, Algorithms, and Networks, I-SPAN 2011, and the 10th IEEE Int. Conf. on Ubiquitous Computing and Communications, IUCC 2011
Y2 - 24 August 2011 through 26 August 2011
ER -