Skip to Navigation Skip to Page Content
With Michela Becchi (center) are Kittisak Sajjapongse, to her left, and Marziyeh Nourian, to her right. Behind them, left to right, are Henry Wu, Ruidong Gu and Vincent Su.

Assistant Professor Michela Becchi, center, is the recipient of a coveted NSF CAREER Award for her research project “Compiler and Runtime Support for Irregular Applications on Many-Core Processors.” Posed here with her is her research team: Kittisak Sajjapongse, to her left, and Marziyeh Nourian, to her right. Behind them, left to right, are Henry Wu, Ruidong Gu and Vincent Su. Photo by Shelby Kardell

Assistant Professor Michela Becchi in MU Engineering’s Electrical and Computer Engineering Department has received the National Science Foundation’s Faculty Early Career Development (CAREER) Award. The prestigious award provides five years of financial support to “junior faculty who exemplify the role of teacher-scholars through outstanding research, excellent education and the integration of education and research within the context of the mission of their organizations.”

Becchi’s awarded project, “Compiler and Runtime Support for Irregular Applications on Many-Core Processors,” aims to facilitate the programming of massively parallel hardware, such as graphics processing units (GPUs), as today’s programmers must be familiar with both parallel programming and the operation of the hardware to get the required performance.

“GPUs offer an unprecedented computing capability on a single board. Unlike desktop computers, which often have two to four processors, GPUs include hundreds to thousands of processor cores,” Becchi explained. “Each processor core is designed to perform a pretty simple computation, but together, they offer a lot of computational power. The difficulty is in the programming.”

GPUs are a natural fit for applications that have a regular structure and can be broken down into simple parts that run the same code, as is the case for image and video processing, for example.

But many other popular applications are more difficult to efficiently run on parallel platforms because they rely on less regular data structures. Social media, Google maps, neuroscience and bioinformatics are examples of applications that rely on network-type structures.

“Network-based applications are more difficult to program,” Becchi said. “Subsets of nodes within the network are processed at different times. The resulting computation follows irregular patterns and involves complex synchronization. Yet, there is inherent parallelism in the processing of large networks, and this is why massively parallel platforms like GPUs may also help these applications.”

The algorithms/software being developed by Becchi’s CAREER project will make it easier for programmers to work with parallel platforms.

Becchi believes that GPUs also have a positive role to play in education.

“Nowadays virtually all computing platforms are parallel, and so there is a strong need to teach students parallel programming,” she said. “Students use GPUs for gaming. Bringing the hardware into the classroom makes it easier to spark their interest.”

Through a contact at Truman State University in Kirksville, Mo., she will organize GPU programming workshops and invite Truman students to MU to expose them to the technology.

Additionally, Becchi has established relationships with a number of IT companies – such as Nvidia, AMD, Intel, NEC and Micron Technology and encourages students to participate in internships with those companies and others.