Overview
Based on the computer architecture, an acceleration library improves the computing efficiency of base software libraries by changing data structures and algorithms of software code and making full use of the high performance of a dedicated chip. Under constraints such as costs, improving the hardware computing capability cannot quickly solve problems. An acceleration library must be used to achieve the optimal hardware performance and provide better services to improve the computing speed.
Functions of an acceleration library:
- For a device with a single-core CPU, if tasks are blocked, the CPU will be in the waiting state, and the computing power of the CPU will be wasted. The multi-thread technology divides time slices to make full use of the CPU capability. This is a software technology.
- For a device with a multi-core CPU, multiple threads need to be scheduled by software to make full use of multiple cores.
- For a device with multiple CPUs, software needs to be used to coordinate CPUs and memory to effectively use resources.
For details on the Kunpeng BoostKit Library, seehttps://www.hikunpeng.com/document/detail/en/kunpengaccel/overview/kunpengaccel.html
Parent topic: Kunpeng BoostKit Library