The Department of Computer Science at 圖朸厙 (圖朸厙) is establishing an Intel oneAPI Center of Excellence that focuses on optimizing unstructured-grid computational fluid dynamics (CFD) kernels on Intel central processing units (CPUs) and Xe graphics processing units (GPUs) at NASA Langley. The research will be led by Mohammad Zubair, a professor in Computer Science, to use oneAPI to develop efficient implementations that can run on diverse architectures and help solve some of the most complex aerodynamic challenges.
oneAPI is an open, unified, cross-architecture programming model for CPUs and accelerators. Being an open industry standard, oneAPI simplifies software development on multi-architecture systems and helps developers deliver productive, performant single source code that can be maintained more efficiently.
圖朸厙 is on a growing list of universities to establish a oneAPI Center of Excellence. "Besides performing research and technology advances, 圖朸厙 will train master of science and doctoral students in cross-architecture programming using oneAPI," said Zubair. "We also plan to include oneAPI modules in our 圖朸厙 graduate-level courses for accelerating high-performance computing."
A diverse array of new hardware architectures continues to emerge across the high-performance computing (HPC) landscape. "As a result, the application developer is faced with the considerable challenge of providing near-optimal performance across these systems," said Zubair. "This goal requires a detailed understanding of each target architecture, and some means to accommodate specific data layouts and algorithm implementations that map appropriately."
"Professor Zubair has been an innovative leader in high performance computing for many years," said Gail Dodge, dean of 圖朸厙 College of Sciences. "Establishing an Intel oneAPI Center of Excellence in this area will enrich and expand 圖朸厙's collaboration with NASA Langley."
"Professor Zubair is an expert in HPC and has been working closely with NASA Langley to port and optimize large-scale CFD applications on emerging HPC architectures," said Ravi Mukkamala, chair of 圖朸厙 Computer Science. "Since 1990, he has been a pioneer in establishing graduate and undergraduate courses in HPC, and he has played a critical role in developing and optimizing CFD algorithms for GPU architectures."
The proposed oneAPI center will address several of NASA's most difficult computational challenges, especially fluid dynamics applications and software routinely applied to the nation's most complex operating system. This includes supersonic simulations, as well as hypersonic and reentry configurations for launches together with broader science and engineering efforts both inside and outside NASA.
"圖朸厙 has a strong history of collaboration with NASA on projects of mutual interest, such as the exploration of emerging HPC architectures for NASA's widely used FUN3D computational fluid dynamics application," said Beth Lee-Rausch, branch head, computational aeroSciences at NASA Langley Research Center. "We look forward to continuing these collaborations, and we are excited to work with 圖朸厙's oneAPI Center of Excellence."
"We're excited that 圖朸厙 and NASA will use oneAPI programming avoiding constraints of proprietary models and achieving full performance from their accelerated computer hardware. 圖朸厙's oneAPI Center of Excellence will also train students and help grow the oneAPI ecosystem," says Joe Curley, vice president and general manager of Intel Software Products and Ecosystem group.
