In this paper, we report the performance of in-house developed parallel staggered grid finite-difference based 2D and 3D seismic acoustic modeling application using PARAM Yuva II. Seismic modeling is used for simulation of seismic wave propagation through earth's subsurface for generation of seismic data. The accuracy of the developed scheme is 4th order in space and 2nd order in time. MPI and OpenMP is used for parallelization. Parallelization model is based upon data decomposition strategy. Different optimization techniques are applied for performance enhancement. For 2D, performance gain of 5.25X is recorded using optimization techniques with respect to the baseline code (MPI + OpenMP). Speedup and efficiency of the application are studied using 16 to 1024 cores and results are presented in the paper. The 2D optimized application is also ported on Xeon Phi co-processor in native mode. The OpenMP scaling is performed on Xeon Phi and comparison with Xeon node results are discussed. 2D and 3D synthetic simple to complex geological models are used for applications outcome demonstration.
[1]
Paulius Micikevicius,et al.
3D finite difference computation on GPUs using CUDA
,
2009,
GPGPU-2.
[2]
R. Madariaga.
Dynamics of an expanding circular fault
,
1976,
Bulletin of the Seismological Society of America.
[3]
T. Kunz,et al.
Three dimensional SEG/EAEG models; an update
,
1996
.
[4]
A. Levander.
Fourth-order finite-difference P-SV seismograms
,
1988
.
[5]
Jeff Linderoth,et al.
Metacomputing and the Master-Worker Paradigm
,
2000
.
[6]
J. Virieux.
P-SV wave propagation in heterogeneous media: Velocity‐stress finite‐difference method
,
1986
.
[7]
Jeroen Tromp,et al.
A perfectly matched layer absorbing boundary condition for the second-order seismic wave equation
,
2003
.
[8]
Jean Virieux,et al.
SH-wave propagation in heterogeneous media: velocity-stress finite-difference method
,
1984
.