Empowering Scientists: Billy Moses’ Compiler Innovations for High-Performance Computing

Access to high performance computing (HPC) is crucial for many scientific fields. Experts in chemistry and physics cannot pursue second PhDs in computer science to gain the coding skills needed for parallel computing.

Billy Moses, a professor at the University of Illinois Urbana-Champaign, received the 2024 SIGHPC Doctoral Dissertation Award for his work titled “Supercharging Programming through Compiler Technology.” He earned his PhD in electrical engineering and computer science at MIT. The SIGHPC Chair praised his dissertation as an important advancement in HPC capabilities.

Moses aims to create tools that allow scientists to use computation without becoming coders. For instance, he collaborates with physicists who simulate black holes. Simple equations in research papers often translate into extensive code involving parallel processing and machine learning. Moses’ tools enable scientists to focus on their expertise while the compiler handles the complexities of HPC.

His research challenges the idea that domain-specific tools are essential for certain tasks. Moses developed the Tapir compiler, which integrates parallelism into the compiler’s structure, facilitating optimizations for parallel code. This work is part of the OpenCilk compiler project. The Enzyme compiler can also automatically generate derivatives from code, improving performance compared to traditional methods. Moses’ approach allows for integration of various compiler tools, enhancing usability.

“Compilers can democratize access to computing,” he states. Instead of requiring users to learn all aspects of computing, his methods let researchers write the code they want while still utilizing advancements like parallel processing and machine learning.

What are the main challenges that scientists face when trying to access high-performance computing resources?

Interview with Billy Moses: Pioneering Access to High-Performance Computing in Scientific Research

News Directory 3: Congratulations on receiving the 2024 SIGHPC Doctoral Dissertation Award for your work titled “Supercharging Programming through Compiler Technology.” How does this recognition feel, and what does it mean for you personally and professionally?

Billy Moses: Thank you! It’s truly an honor. Receiving this award not only validates my research but also highlights the importance of making high-performance computing (HPC) more accessible to scientists across various disciplines. Personally, it’s rewarding to see that my work is being acknowledged as a significant advancement in the field, reinforcing my commitment to bridge the gap between programming and scientific inquiry.

ND3: You mentioned the challenge scientists face in gaining the necessary coding skills for HPC. Can you elaborate on why access to HPC is crucial for fields like chemistry and physics?

Moses: Absolutely. High-performance computing is essential for simulating complex phenomena that cannot be easily replicated in a laboratory setting. In fields like chemistry and physics, researchers often deal with intricate models—such as molecular interactions or astrophysical events—that require vast amounts of data processing. Without HPC, the scope and depth of research are severely limited. However, the steep learning curve of programming can deter scientists who are not trained in computer science.

ND3: Your dissertation focuses on developing tools that enable scientists to leverage HPC without extensive programming knowledge. How specifically do you envision this materializing in practical applications?

Moses: My goal is to create user-friendly tools powered by advanced compiler technologies that simplify the programming process. For example, when physicists simulate black holes, they work with equations that, if translated linearly into code, require lengthy and complex programming. I aim to develop solutions that allow researchers to input equations or define parameters in a straightforward manner, with our tools automatically optimizing the code for efficient execution on HPC resources. This means scientists can focus on their research rather than getting bogged down in coding challenges.

ND3: Collaborating with physicists and other scientists is a significant part of your work. How do these collaborations influence the development of your tools?

Moses: Collaboration is key. By working alongside physicists, chemists, and other domain experts, I gain valuable insights into their specific needs and challenges regarding HPC. Their feedback informs the design of our tools, ensuring they fit seamlessly into existing workflows. Furthermore, these partnerships help to identify common barriers scientists face in computing, allowing us to create solutions that are genuinely impactful and user-centric.

ND3: What do you think the future holds for HPC in scientific research, especially for those who lack advanced coding skills?

Moses: I believe we are on the brink of a paradigm shift. As more researchers understand the potential of HPC but struggle with coding, there will be an increasing push for intuitive tools that democratize access to these technologies. The future of HPC will likely involve more integration of machine learning and AI to personalize computing experiences, making high-level computational tools more accessible to scientists across various disciplines. If we can empower more researchers to utilize HPC effectively, the potential for groundbreaking discoveries is tremendous.

ND3: Lastly, do you have any advice for young scientists who aspire to make an impact in HPC or related fields?

Moses: Absolutely. My advice would be to stay curious and collaborative. Even if you don’t have a background in computer science, there are numerous resources available today to learn the basics of programming and computation. Engage with interdisciplinary teams, seek mentorship, and always be open to learning from those outside your field. The intersection of disciplines is where innovation often occurs, and harnessing the power of HPC will require diverse perspectives and skills.

ND3: Thank you, Billy. Your insights are both enlightening and inspiring. We wish you continued success in your work!

Moses: Thank you! It’s been a pleasure discussing these vital topics. I’m excited about the future of HPC and its potential to transform scientific research.

—

For more updates and insights from the forefront of science and technology, be sure to check out News Directory 3.

Moses’ work eliminates the need for separate tools for different tasks. His research allows experts in various fields to use their knowledge within a unified framework, optimizing processes across disciplines. For example, astrophysicists can compute equations while simultaneously leveraging parallel processing capabilities.

With his team, Moses published a paper showing how combining optimization techniques can yield greater improvements. Their collaborative work led to a Best Student Paper Award at SC22.

Moses believes his dissertation is just the start. He sees potential for further development, especially in communication and other domains. He leads the PRONTO lab at Illinois, where his team works on distributed computing and numerics.

Billy Moses is a professor at the Grainger School of Engineering at the University of Illinois. He also works with the electrical and computer engineering and Coordinated Science Laboratory departments.