Abstract
Nowadays, it is very common that a processor includes a SIMD (Single Instruction Multiple Data) unit in order to accelerate application processing. While a SIMD unit is a part of a processor, it evolves more rapidly than the integer unit of the processor. Since the use of an FPU (Floating Point Unit) and a SIMD unit is basically abandoned from the kernel, there can be places inside the kernel where a SMID unit works effectively to deal with a large amount of data processing. This paper describes our preliminary work to explore the possibility to utilize a SIMD unit in the kernel. We performed preliminary experiments by using UML (User Mode Linux) and show that data copying can be improved.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have