NETL researchers recently took a significant step forward in harnessing the power of the world’s largest computer chip — the Wafer-Scale Engine (WSE) — by using an application programming interface designed in-house to connect commercial computational fluid dynamics (CFD) software through data-file sharing with the next-generation computing technology.

CFD software enables researchers to model the complex behavior of energy systems to help increase efficiency and performance. Typically, CFD relies on high-performance supercomputing, but such systems based on the WSE hold the promise to run high-fidelity simulations hundreds of times faster and with far less energy.

“While the WSE is an incredibly powerful computer chip, it’s architecture requires a new programming interface to integrate with any scientific software,” said NETL’s Dirk Van Essendelft. “We have been working with Cerebras Systems, the creator of the WSE, to develop the Wafer-scale engine Field equation Application programming interface (WFA) to achieve this integration.” 

In this work, the NETL team used the WFA on Carnegie Mellon’s Neocortex computational resource, which features two of Cerebras’ CS-2 WSE-enabled systems. The researchers generated a CFD case data file for 3D lid driven cavity flow in the commercial software ANSYS Fluent, processed the file using the WFA, and placing data onto Neocortex, used the data to calculate a steady state solution, then assembled the resulting data into a file that ANSYS Fluent could interpret.

“The test case we used is a simple one, but it does represent a commercial software integration on the WSE,” Van Essendelft said. “Tapping into the power of the WSE may allow us to solve more complex scientific problems in the future, which could lead to breakthroughs in the way energy systems are designed.”

The software integration is the latest in several successes the team has achieved in relation to research using the WSE. The NETL team was the first to work with Cerebras to demonstrate a non-artificial intelligence high-performance computing application on the WSE. They were also the first to develop a high-level computing language for the WSE and demonstrate and characterize the first engineering application.

Using the WSE, NETL researchers have also explored coupled, single-phase CFD; made advancements using the Ising model, a cornerstone in statistical physics and a valuable tool for evaluating the performance of emerging computer hardware; nearly completed a molecular dynamics code; and found recent successes in DOE’s Science-informed Machine Learning for Accelerating Real-Time (SMART) Decisions in Subsurface Applications initiative in the area of finite volume, multiphase, compressible reservoir modeling.

“Through partnerships with entities such as Cerebras Systems, Carnegie Mellon and others, we are exploring new and powerful uses for the WSE,” Van Essendelft said. “While the architecture was designed and is primarily used to tackle difficult artificial intelligence problems, ongoing research is proving that the chip also works remarkably well for other scientific problems.”

This work was funded and conducted under the “Novel Field Equation Integrator” project as part of the U.S. Department of Energy’s Laboratory Directed Research and Development (LDRD) Program, which allocates discretionary funding to national labs for high-risk, high-reward projects. LDRD funds let NETL explore forefront areas of science and technology, serve as a proving ground for new concepts, seed future research programs, and attract and retain high-caliber scientists and engineers.

NETL is a DOE national laboratory that drives innovation and delivers solutions for an environmentally sustainable and prosperous energy future. By using its world-class talent and research facilities, NETL is ensuring affordable, abundant and reliable energy that drives a robust economy and national security, while developing technologies to manage carbon across the full life cycle, enabling environmental sustainability for all Americans.