Towards Exascale Geophysical Applications
|Convenors:||M. Hanzich (Barcelona Supercomputing Center)|
|A. St-Cyr. (Shell)|
|G. Roth (Nvidia)|
|P. Thierry (Intel)|
|Date:||Wednesday 1 June|
The quest for oil & gas resources is directly driven by the energy demands of the ever growing population of the Earth.
Producing existing resources is not enough to respond to the demand and exploration Geophysics needs to discover new hydrocarbons using more physics and more complex numerical algorithms. The latter have many stages relying on High Performance Computing (HPC) capabilities.
For instance, HPC is not only applied in the final steps of imaging or migration of the data, but is also involved at the pre- and post-process phases.
All phases need to prepare huge amounts of data for processing and the interpretation of the results requires specialized tools to extract the relevant information.
In HPC, there is an emerging trend towards the integration of the ever-evolving hardware with software that incorporates heterogeneous solutions for computation, visualization, storage, interconnect, energy consumption, to name a few. Of course, this increases the complexity of the development environment for supporting exploration geophysics. However this is necessary in order to exploit the upcoming Exascale platforms to their full extent.
Transparent APIs or compilers for supporting parallelism, reliable measuring tools and methodologies for performance evaluation are only some of the mechanisms that should be in place to cope with the exploration Geophysics needs. At the end, reaching high-performance in Geophysics must remain achievable for every end user or dedicated developer regardless the complexity of underlying environment.
The aim of this dedicated session will be to review the current and future state of HPC solutions for many different geophysical problems such as modelling, migration or inversion and others.
This includes all steps needed to process exploration data including data preprocessing, the mining and visualization of results. Of course, these solutions need to be aware of the characteristics of modern, maybe heterogeneous, hardware architectures and of the latest advances in algorithmic, programming models, languages and evaluation tools and methodologies.