HySEA Tsunami Simulation Aided by SYCL*

Rob_Mueller-Albrecht · ‎12-21-2023

Researchers at the Differential Equations, Numerical Analysis and Applications (EDANYA) Group of the University of Málaga have worked with software engineers at Intel Corporation to enable their Tsunami simulation workload for SYCL offload, freeing it from vendor-lock to a specific GPU. They thus achieved that this simulation framework can run on various hardware configurations. This allows for wider deployment, driving our understanding of earthquakes and their interactions with water and coastal regions forward.

They research various numerical simulation use cases looking at geophysical fluids employing the numerical solution of hydrodynamic models. Examples include:

Numerical Analysis of Systems of Partial Differential Equations:

Finite Element Methods
Finite Volume Methods

Geophysical Flows Simulation:

Shallow-Water models
Sediment transport
Submarine and aerial avalanches
Tsunami generation, propagation, and inland penetration
Oceanographic applications
Hydrodynamic-Biology coupled models
Pollutant transport
Hazard assessment (dambreaks, avalanches, floods, ...)

Some of this work is done with the help of HySEA (Hyperbolic Systems and Efficient Algorithms), a high-performance package developed by the EDANYA group at the University of Málaga, Spain, to simulate geophysical flows.

Tsunami-HySEA specifically is a codebase for the simulation of Tsunamis triggered by earthquakes and also developed by the EDENYA Group Research. It is based on the Finite Volume Method and was originally optimized for use with CUDA and NVIDIA GPUs. It is now available with full SYCL support.

"Migrating our tsunami simulation and prediction code 'Tsunami HySEA' from CUDA to SYCL was a decision that allowed our users and us to be independent of individual hardware vendors. Previously, the application required NVIDIA* GPUs. Now, with just one source code, the application runs with a similar performance on NVIDIA hardware, but can also run on other vendors' CPUs and GPUs, such as Intel® Data Center GPU Max Series. With a very good performance, its ability to run on different platforms, and a seamless transition process, SYCL has become an extremely useful tool for our parallel computing efforts. We recommend SYCL to any organization that wants to realize the full potential of heterogeneous computing and achieve excellent performance on multiple platforms using a single code.”

—EDANYA Group at University of Málaga

You can access the TsunamiHySEA SYCL version source code here.

Understanding the Physics of Natural Disasters

A tsunami is a series of water waves triggered by the displacement of a large volume of water, like an ocean or a large lake. Although we commonly think of underwater or near-water earthquakes as the most common root cause, other events like volcanic eruptions, landslides, or even meteorite impacts could cause similar displacements.

Seismic tsunamis can be generated when the sea floor abruptly deforms and vertically displaces the overlying water. Tectonic earthquakes are a particular kind of earthquake that are associated with the earth's crust deformation. When these earthquakes occur beneath the sea, the water above the deformed area is displaced. Thus, a tsunami can be triggered when thrust faults associated with convergent or destructive plate boundaries move abruptly, resulting in vertical water displacement. Movement on extensional faults can also cause seabed displacement, but only the largest of such events cause enough displacement to give rise to a significant tsunami.

The resulting tsunamis generally consist of a series of waves that tend to have a small wave height offshore and a very long wavelength - often hundreds of miles long. In contrast, typical ocean waves have a wavelength of only 30 or 40 yards. Although the visible impact of tsunamis is limited to coastal areas, their destructive power can be enormous and have a serious impact on the seabed as well.

The impact on coastal regions is mostly associated with wave shoaling. As tsunami waves enter shallower water, wave height increases as the group velocity of the wave decreases. In other words, looking at it from an energy conservation perspective, as the wave propagation is slowed down, it tries to maintain the flow of energy constant by increasing energy density, resulting in higher amplitude. As waves approach the shore and the water gets shallower, the waves get taller, slow down, and get closer together.

This picture gets complicated by overlayed processes like refraction, diffraction, reflection, and wave breaking as it moves onshore, adding to its destructive impact. The result is a series of waves pushing ashore like a fast-rising tide extending far inland and exhibiting powerful currents. If a tsunami-causing disturbance occurs close to the coastline, a resulting tsunami can reach coastal communities within minutes.

In short, it is vital not only for scientific understanding but also for emergency response planning that we have reliable models that can simulate the propagation of a tsunami wave and assess its possible impact reliably and accurately.

The Power of Simulation and Prediction

This is where HySEA comes in: a family of geophysical codes based on either single-layer, two-layer stratified systems, or multilayer shallow water models developed for the simulation of geophysical flows, including tsunamis generated by earthquakes or landslides, river floodings, sediment transport, turbidity currents and more.

Tsunami-HySEA is the numerical model in this family of simulation implementations targeting earthquake-generated tsunamis. Taking advantage of a GPU for accelerated offload computation allows for a faster-than-real-time (FTRT) implementation. Leveraging the real-time measurements of comprehensive system models allows for evaluating the full geophysical system at faster speeds than in real-time, thus enabling the simulation to predict the whole system's behavior efficiently. Tsunami-HySEA looks at the end-to-end evolution of a tsunami wave from generation and propagation to coastal impact.

Okada's fault deformation model [1] is used to predict the initial bottom deformation transmitted instantaneously to the sea surface, generating the tsunami wave. This method assumes an earthquake can be considered a single-fault plane rupture. This fault is described by a series of parameters comprising dip angle, strike angle, rake angle, fault width, fault length, and fault depth. Tsunami-HySEA can combine several fault planes to model the complete seafloor deformation. Moreover, each fault could be applied at different time steps to simulate the full rupture time.

The computational demand on the platform configuration used to run these simulations is significant. The physics governing the tsunami propagation are based on hyperbolic systems of partial differential equations for the interaction between the waves and the conservation laws that govern it. Numerical solutions of these partial differential equations, as expressed in the finite volume method (FVM) and hyperbolic shallow water equations, form the basis for this type of simulation. This gets additionally complicated by the large wavelength of tsunami waves requiring a large geographic area to be modeled, while many smaller-scale complex interactions influence the coastal inundation phase. This disparate spatial scaling of factors influencing the tsunami’s behavior only increases the need for computing power and precise modeling.

SYCL Opens Up Deployment on Diverse Hardware

Porting the Tsunami-HySEA code base to support SYCL moves it to an open-source and open-standards-based programming model supported by Khronos Group as well as the Unified Acceleration (UXL) Foundation drive towards an open standard accelerator ecosystem.

Reliable and accurate simulation and, thus, prediction of seismic tsunamis is now possible on a wide range of CPU and accelerator combinations, allowing it to be used and deployed at more locations.

This will help scientists predict tsunami impact for more coastal regions, giving emergency response teams more precise data to plan for the day it happens.

Become part of the effort to make high-performance cross-architecture compute transparent, portable, and flexible. Include SYCL as the accelerator and GPU offload solution in your code path. Adopt oneAPI as the means to implementations free from vendor lock.

Get started with HySEA-Tsunami’s source repository.
Download the latest HySEA releases.

The Intel® DPC++ Compatibility Tool and the CUDA to C++ with SYCL Migration Portal are the convenient starting point for your own migration to SYCL.

Make SYCL part of your software solution.

SYCL Resources

HySEA Resources

[1] Okada, Y. (1985) Surface Deformation due to Shear and Tensile Faults in a Half-Space. Bulletin of the Seismological Society of America, 75, 1135-1154.