当前位置：网站首页>Introduction to arm Cortex-M learning

Introduction to arm Cortex-M learning

2022-06-13 02:12:00 【Code changes the world CTW】

Quick links :
.
Personal blog notes guide Directory ( All )

brief introduction

Cortex-M The processor family is based on M-Profile framework , It can provide low latency and highly deterministic operation for deep embedded systems . Our latest generation Cortex-M The processor is Cortex-M55.Cortex-M55 yes The first paragraph be based on Armv8.1-M Architecture's processor , Use vector processing to extend Arm Helium technology .Cortex-M55 For the next generation of small embedded devices （ Including wearable devices 、 Smart speakers, etc ） Bring a higher level of machine learning and signal processing performance .

What is? Arm Helium technology ？

Arm Helium The technology is Arm Cortex-M Processor family M-Profile Vector Extension (MVE).Helium yes Armv8.1-M The extension of Architecture , It can provide significant performance improvement for machine learning and digital signal processing applications of small embedded devices .

Helium Is an optional extension , for Armv8.1-M The processor provides higher signal processing capability . Cortex-M55 Is the first to support this technology Arm processor .

Add more than 150 New scalar and vector instructions , There are many implementation options ：

Helium option omitted—— With optional scalar FPU Of Armv8.1-M Integer kernel （ Double precision support is also optional ）.
With optional scalars only FPU The integer of （ Double precision support is also optional ）.
With scalar FPU（ Double precision support is also optional ） The integer of + floating-point Helium（ Support vector single precision and semi precision ）.

Integers Helium Support efficient computing 8 position 、16 Bit and 32 Bit point data .16 Bit and 32 Bit fixed point format is widely used in traditional signal processing applications , For example, audio processing .8 Bit point format for machine learning processing （ For example, neural network computing ） And image processing can be important .

Helium Floating point data types are also supported , Including single precision floating-point numbers （32 position ） And semi precision floating point numbers （16 position ）.