Scientists have devised a method to roughly double the processing speed of PCs and smartphones without upgrading hardware and halve power consumption
In general, increasing the processing performance of a computer requires upgrading the hardware itself or improving the computer's operating environment. Researchers at
Simultaneous and Heterogenous Multithreading | Proceedings of the 56th Annual IEEE/ACM International Symposium on Microarchitecture
https://dl.acm.org/doi/10.1145/3613424.3614285
Method identified to double computer processing speeds | UCR News | UC Riverside
https://news.ucr.edu/articles/2024/02/21/method-identified-double-computer-processing-speeds
Radical New Discovery Could Double The Speed of Existing Computers : ScienceAlert
https://www.sciencealert.com/radical-new-discovery-could-double-the-speed-of-existing-computers
In addition to a CPU that performs general processing, modern computer devices are equipped with multiple processors and hardware accelerators, such as GPUs that mainly handle graphics processing and tensor processing units (TPUs) for machine learning. In many cases, it can be said that it is essentially a heterogeneous environment.
Traditional programming models focus on using only the most efficient processing units in each area of code. This certainly makes sense, since the code that runs a spreadsheet program and the code that runs a generated AI text bot are not handled in the same way. However, with the method of processing tasks individually with a specific processor, a bottleneck occurs when transferring information to the next processing unit after performing one process.
To take full advantage of the strengths of heterogeneous computers, Hung-Wei Tseng , an associate professor of electrical and computer engineering at the University of California, Riverside, and his colleagues have developed a 'simultaneous' execution model that executes parallel tasks on multiple processors simultaneously. and heterogenous multithreading (simultaneous heterogeneous multithreading: SHMT).
In contrast to conventional models, SHMT allows the simultaneous use of different types of processing units for the same code region. Additionally, Tseng et al. claim that SHMT provides an abstraction and runtime system to facilitate parallel execution of code.
To test SHMT, Tseng and colleagues ran sample code on a computer system containing an ARM Cortex-A57 CPU, NVIDIA GPU, and Google Edge TPU. As a result, SHMT increased the execution speed of the sample code by approximately 1.96 times and reduced energy consumption by 51%.
By using existing processing components at the same time, you can not only save on hardware costs, but also reduce your carbon footprint due to energy consumption. Tseng commented, ``If you already have a processor, there is no need to add a new processor.''
In addition, SHMT is still in the stage of testing whether the idea has potential, and it is not possible to immediately introduce the proposed system on smartphones and smart watches.
Related Posts:
