Oil and Gas

	RELATED PAPERS »FD modeling beyond 70Hz with FPGA acceleration »Surviving the End of Scaling of Traditional Microprocessors in HPC »Fast 3D ZO CRS Stack »Anisotropic reverse-time migration using co-processors »An Implementation of the Acoustic Wave Equation on FPGAs »Accelerating Subsurface Offset Gathers for 3D Seismic Applications

With total world petroleum consumption approaching 100 million barrels a day, discovery of new oil and gas reserves is more important than ever. New software tools can have a significant impact on seismic data interpretation for exploration but at a considerable computational cost. Maxeler's technology can be deployed to deliver the next generation of algorithms at a cost comparable to lower fidelity approaches, as well as to increase throughput and decrease turnaround times for current applications.

With up to 192GB of dataflow compute memory in a dense 1U form factor, Maxeler high-performance dataflow hardware solutions are ideal for processing large seismic datasets without disk or network I/O bottlenecks. High bandwidth to dataflow compute memory and Maxeler domain-specific data encoding and streaming data (de)compression can enable significant speed-up of algorithms that are typically regarded as memory bound, such as sparse matrix solvers.

For Forward Modeling, Reverse Time Migration and Full Waveform Inversion the high-level MaxGenFD domain-specific compiler allows geoscientists to write a high level description of a finite difference propagator and easily create highly optimized dataflow implementations, including automatic parallelization across multiple dataflow compute cards and multiple compute nodes.

We are collaborating with the Center for Earth and Environmental Science (CEES) at Stanford University to accelerate key algorithms in the seismic computing domain. Pushing forward the frontiers of computational geoscience, CEES is engaging computer scientists and architects to design software and hardware targeted to solving Earth and environmental science problems.