|
Titel |
Spectral-element seismic wave propagation on emerging HPC architectures |
VerfasserIn |
Daniel Peter, Qiancheng Liu, Dimitri Komatitsch |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250151807
|
Publikation (Nr.) |
EGU/EGU2017-16567.pdf |
|
|
|
Zusammenfassung |
Seismic tomography is the most prominent approach to infer physical properties of Earth’s
internal structures such as compressional- and shear-wave speeds, anisotropy and
attenuation. Using seismic signals from ground-motion records, recent advances in
full-waveform inversions require increasingly accurate simulations of seismic wave
propagation in complex 3D media to provide access to the complete 3D seismic
wavefield. However, such numerical simulations are computationally expensive and need
high-performance computing (HPC) facilities for further improving the current state of
knowledge.
During recent years, new multi- and many-core architectures such as graphics processing
units (GPUs) have been added to available large HPC systems. GPU-accelerated computing
together with advances in multi-core central processing units (CPUs) can greatly accelerate
scientific applications. To employ a wide variety of hardware accelerators for seismic wave
propagation simulations, we incorporated a code generation tool BOAST into an
existing spectral-element code package SPECFEM3D_GLOBE. This allows us to use
meta-programming of computational kernels and generate optimized source code for both
CUDA and OpenCL languages, running simulations on either CUDA or OpenCL hardware
accelerators. We show here benchmark applications of seismic wave propagation on
GPUs and CPUs, comparing performances on emerging hardware architectures. |
|
|
|
|
|