Imagine being able to increase the processing power of your existing devices, like smartphones, tablets, or computers, without having to add new hardware.
Hung-Wei Tseng, an associate professor at UC Riverside, has developed a new approach to computer architecture that he believes can do just that. In a recent paper titled, “Simultaneous and Heterogeneous Multithreading,” Tseng introduces a new method for using the existing components in modern devices more efficiently.
How It Works
Many devices today already have separate components for tasks like graphics processing, artificial intelligence, and digital signal processing. Tseng’s new method, called “Simultaneous and Heterogeneous Multithreading” or SHMT, allows these components to work together simultaneously, without creating a bottleneck.
In their research, Tseng and UCR computer science graduate student Kuan-Chieh Hsu developed a proposed SHMT framework on an embedded system platform that uses a multi-core ARM processor, an NVIDIA GPU, and a Tensor Processing Unit hardware accelerator at the same time. The results were impressive, with a nearly 2 times speedup and a 51% reduction in energy consumption.
The Benefits
The ability to use existing processing components more effectively could lead to cost savings for hardware, as well as a reduction in energy consumption and carbon emissions from large data processing centers. It could also help conserve scarce freshwater used to keep servers cool.
The Future
While Tseng’s research shows promising results, he acknowledges that further investigation is needed to address various questions about implementation, hardware support, code optimization, and which applications could benefit the most. However, the paper has already gained recognition from IEEE and will be published in their “Top Picks from the Computer Architecture Conferences” issue.