Imagine doubling the processing power of your smartphone, tablet, personal computer, or server using the existing hardware already in these devices.
Hung-Wei Tseng, a UC Riverside associate professor of electrical and computer engineering, has introduced a new concept in computer architecture to achieve just that. In a recent paper titled, “Simultaneous and Heterogeneous Multithreading,” Tseng outlines a paradigm shift in maximizing the processing power of various devices.
Tseng points out that modern computer devices come equipped with GPUs, AI and ML accelerators, or DSP units as integral components. However, these components often operate independently, leading to a bottleneck in information processing.
In their paper, Tseng and UCR computer science graduate student Kuan-Chieh Hsu propose the concept of “simultaneous and heterogeneous multithreading” (SHMT). They have developed a SHMT framework on an embedded system platform that utilizes a multi-core ARM processor, an NVIDIA GPU, and a Tensor Processing Unit hardware accelerator simultaneously.
Through this approach, the system achieved a 1.96 times speedup and a 51% reduction in energy consumption.
“You don’t need to add new processors because the existing ones can be utilized effectively,” Tseng explained.
This innovative approach could lead to significant cost savings in computer hardware and a reduction in carbon emissions from energy consumption in data processing centers. It also has the potential to decrease the demand for freshwater needed to cool servers.
While Tseng’s paper highlights the promising implications of SHMT, further research is necessary to address various implementation challenges, hardware compatibility, code optimization, and the optimal applications for this technology.
The paper was presented at the 56th Annual IEEE/ACM International Symposium on Microarchitecture in Toronto, Canada, where it received recognition from the IEEE. It was selected as one of the 12 papers to be featured in the group’s “Top Picks from the Computer Architecture Conferences” issue scheduled for publication in the upcoming summer.
