An innovative approach to the use of processors opens up new opportunities.
A recent study from the University of California, Riverside (UCR) presented a groundbreaking discovery that promises to significantly speed up the performance of modern computers without the need for hardware improvements. The essence of the discovery is to use more efficient software algorithms that can double the speed of devices.
Researchers have developed a method called simultaneous and heterogeneous multithreading (SHMT), which allows multiple processors to be used in parallel in phones, computers, and other gadgets. This approach can not only increase hardware performance, but also significantly reduce power consumption.
A key feature of SHMT is that it uses pre - existing processors in devices - the central processing unit (CPU), graphics processing unit (GPU), and tensor processing unit (TPU) - to process tasks in parallel. This allows you to avoid slowdowns caused by data transfer between different blocks and increase the efficiency of program execution.
During testing, which included the use of an ARM Cortex-A57 processor, a GPU from Nvidia and Edge TPU from Google, the execution of sample code using SHMT showed acceleration of 1.95 times and a reduction in power consumption by 51%.
Despite the significant potential of the new technology, researchers emphasize that before its widespread adoption, it is necessary to solve a number of problems related to the distribution of computing operations between different types of processors and their subsequent synchronization.
The discovery was presented at the 56th annual IEEE / ACM International Symposium on Microarchitecture in Toronto, Canada, and has already generated a lot of interest in the scientific community, waiting for further research and development in this area.
A recent study from the University of California, Riverside (UCR) presented a groundbreaking discovery that promises to significantly speed up the performance of modern computers without the need for hardware improvements. The essence of the discovery is to use more efficient software algorithms that can double the speed of devices.
Researchers have developed a method called simultaneous and heterogeneous multithreading (SHMT), which allows multiple processors to be used in parallel in phones, computers, and other gadgets. This approach can not only increase hardware performance, but also significantly reduce power consumption.
A key feature of SHMT is that it uses pre - existing processors in devices - the central processing unit (CPU), graphics processing unit (GPU), and tensor processing unit (TPU) - to process tasks in parallel. This allows you to avoid slowdowns caused by data transfer between different blocks and increase the efficiency of program execution.
During testing, which included the use of an ARM Cortex-A57 processor, a GPU from Nvidia and Edge TPU from Google, the execution of sample code using SHMT showed acceleration of 1.95 times and a reduction in power consumption by 51%.
Despite the significant potential of the new technology, researchers emphasize that before its widespread adoption, it is necessary to solve a number of problems related to the distribution of computing operations between different types of processors and their subsequent synchronization.
The discovery was presented at the 56th annual IEEE / ACM International Symposium on Microarchitecture in Toronto, Canada, and has already generated a lot of interest in the scientific community, waiting for further research and development in this area.
