A recent study conducted by the University of California in Riverside (UCR) has unveiled an innovative discovery with the potential to greatly enhance the performance of modern computers without the need for hardware upgrades. The breakthrough involves the implementation of more efficient software algorithms that can effectively double the speed of devices.
Researchers at UCR have introduced a method known as simultaneous and heterogeneous multipro-flow (SHMT) execution, which enables the simultaneous utilization of multiple processors in various electronic devices such as phones and computers. This approach not only boosts device productivity but also significantly reduces power consumption.
One of the distinctive features of SHMT is its utilization of existing processors in devices – the central processor (CPU), graphic processor (GPU), and tensor processing unit (TPU) – for parallel task processing. By eliminating the need to transfer data between different blocks, SHMT enhances program efficiency and eliminates potential slowdowns.
Through testing involving the ARM Cortex-A57 processor, NVIDIA GPU, and Google Edge TPU, the execution of sample code using SHMT demonstrated a 1.95-fold acceleration and a 51% reduction in energy consumption.
Despite the promising capabilities of this new technology, researchers underscore the need to address various challenges related to distributing computational operations among different processor types and ensuring their synchronization before widespread implementation can occur.
This groundbreaking discovery was unveiled at the 56th annual IEEE/ACM International Symposium on Microarchitecture in Toronto, Canada. It has already captured the attention of the scientific community, sparking anticipation for further research and advancements in this domain.