Accelerated Computing Platform Lab
Gwangsun Kim, Hyojin Sung, and Eunhyeok Park (Computer Science and Engineering)
The Association of Computing Machinery of the US, one of the most prestigious societies in the computer industry, defines ‘computing’ as any and all processes performed through the technology of computers, which can range anywhere from simple computer-aided calculations to highly complex data processing. Demands for a better-performing computing system are on the rise as the global production of data increases exponentially. According to IBM of the US, we as a planet, churned out 2.5 quintillion (1 quintillion equals 1 followed by 18 zeros) bytes of data every day as of 2017. To even hear the word ‘quintillion’ eludes us from deciphering the true magnitude of the quantity it contains. The key to performance improvement lies in understanding the components of the computing system – algorithms, software and hardware – and in the creation of designs that efficiently integrate these components.
The Accelerated Computing Platform Lab, jointly directed by Professors of Gwangsun Kim, Hyojin Sung, and Eunhyeok Park at the Department of Computer Science and Engineering, POSTECH, aims to deliver superior computing performance. Each professor conducts optimization research in his/her own specialized field before coming together to create integrative solutions. Professors Kim, Sung and Park are responsible for hardware, software and algorithm optimization respectively. The fact that one single lab handles all three components that determine computing performance improvement is quite a rarity, even from a global perspective.
In a computing system, hardware is represented by Central Processing Units or CPUs that play an essential role in computing power. In addition to CPUs, a variety of hardware components have emerged, including Graphic Processing Units (GPUs), accelerating chips, and Neural Processing Units (NPUs) which are mounted on smartphones. In tandem with the increasing performance of AI, the structure of hardware that deploys and optimizes algorithms continues to evolve. The increasing diversity of hardware requires software technology to adapt accordingly. Different kinds of hardware demand different compilers that translate man-made algorithmic commands into machine codes readily understood by the hardware.
The three professors pursue optimization in their own specialized fields. “Optimization is not special; it is simply part of our innate human instinct. In our everyday lives, we are continuously and unconsciously searching for ways in which we can use energy more efficiently and generate superior results”, Professor Sung noted. “We are optimizing algorithms, with a particular focus on improving the performance of AI. As hardware, software and algorithms are optimized, overall efficiency naturally increases, which is precisely what makes optimization so appealing”, Professor Park added. “Let’s take semiconductor processing technology as an example: if we optimize each component of this technology and then put these components together, the optimization of just this act alone could increase the efficiency more than a hundred fold, even without semiconductor technology advancing any further”, Professor Kim asserted.
The ACP Lab, created through mutual understanding among three like-minded professors, predicts that the computing system will continuously evolve to meet whatever needs the time calls for. “Our research area is particularly intriguing as we are in the midst of witnessing a number of dynamic developments in the computing system sector and the possibilities in the years ahead are essentially without limit”, Professor Kim noted, adding “What we do is fundamentally rooted in system research, and yet we all bring to the table our own unique expertise from various fields. Ultimately, this is what will position the ACP Lab to produce the best-possible optimization outcomes to keep pace with the ever-evolving computing system”.
Head of Lab
Institute of Artificial Intelligence 421