Researchers at Lobachevsky University of Nizhny Novgorod (UNN) published their study on the more efficient use of RISC-V microprocessor resources in the Lobachevskii Journal of Mathematics. This method helps to speed up the execution of certain algorithms in the OpenCV library – the foundation of "computer vision" for robotics, medicine, mobile photography, and autopilot development – by tens of percent.
Thanks to the Novosibirsk method, during vector calculations, the microprocessor performs the same operation on several "portions" of data in parallel during the execution of one instruction. This significantly speeds up calculations.
When using a smartphone, we expect all operations to be performed very quickly, on the fly, without pauses or freezes. Meanwhile, the tasks we entrust to mobile devices are becoming more and more complex. Therefore, software developers try to write programs in such a way as to make the best use of the processor and memory subsystem capabilities so that they ultimately work quickly. There is hope that we may soon see RISC-V processors in mobile and server devices.
According to the university, the research is supported by the "Priority-2030" program of the Russian national project "Science and Universities."
Read more on the topic:
Hundreds of times faster than neural networks: Russia to launch experimental photonic processor
First Russian high-precision superconducting quantum processor starts operating
New Russian method developed for fast three-qubit operation on fluxoniums