Illinois researchers create method to simplify programming for clusters of computers
The problems scientists try to solve are becoming more complex every day, requiring more computer power and intricate data analysis. As we face greater societal challenges, such as large-scale optimization, engineering simulation, and artificial intelligence, researchers at Illinois are building systems that will help make these problems easier to solve.
Martin Wong, CSL professor of electrical and computer engineering and executive associate dean of the College of Engineering, and Tsung-Wei Huang, who just received his PhD in electrical and computer engineering, received a $400,000 grant from the National Science Foundation to develop software to simplify the programming for computer clusters.
Today, it’s easy to connect many computers through the Internet, allowing data to pass among computers. However, programming these computer clusters to deliver correct functionalities is extremely difficult, and Wong and Huang saw a need for a new parallel programming model.
Wong and Huang are developing a system, called DtCraft, that will automatically deliver data correctly, which will allow researchers to focus on their research problem, not how the computers interact.
“Before we had these tools, programmers would have to code in all the details about how the computers connected and sent data to one another themselves,” said Wong. “Now, this program does a lot of the work for programmers, figuring out how to best transfer data, so the programmers only have to work with it on a high level.”
Easing this burden on researchers allows them to focus on their work, knowing that the code behind-the-scenes is being handled by the program.
Wong and Huang have applied the system to solve the circuit timing analysis problem, a vital stage to overall semiconductor design flow. They developed a tool called OpenTimer, an award-winning timing analysis tool in ACM TAU Contests 2014-2017. It is also the basis of Huang’s PhD dissertation.
With DtCraft, OpenTimer can easily leverage the power of computer clusters to efficiently analyze the timing of modern circuits that incorporate billions of transistors.
“Correct timing analysis is critical for building complex chips for complex systems. Chips are everywhere, and they’re getting larger and more complex every day,” said Wong. “In order to make them correctly, you need this kind of software to make sure you design chips with the right functionality. We’re hoping this system helps do that.”
The grant supports the ongoing efforts to strengthen both OpenTimer and DtCraft to assist with large-scale timing analysis using the power of computer clusters. OpenTimer is open-source, meaning anyone can download it and use it for their research.
“We want researchers to use it and to build on top of the code, so the field can progress faster,” said Huang. “The more students we engage with the program, the more we can advance the field and make an impact.”