mmemcpy

memcpy() written in ARM Cortex-M Assembly. Designed for Cortex-M4 and M0+ cores. Probably works on other ARMv6-M and ARMv7-M architectures.

Status

Does everything I need it to. Will fix bugs as I come across them.

How to use this in an embedded project

1. Copy mmemcpy.s from the common/src folder into your project
2. Add a function declaration for memcpy_() somewhere in your project. The function declaration should look something like extern void* memcpy_(void *destination, const void *source, size_t num);.
3. Use memcpy_() just like you would the normal memcpy() function.
4. Done!

How to measure the size of this project

1. copy these files into an embedded project.
2. compile the project.
3. measure the size of the mmemcpy.o object file using size -A.
4. the total size is found from summing all relevant sections together.

Details

Compiles down to 132 bytes on arm-none-eabi-gcc. mmemcpy.s is about 80 lines of code (per David A. Wheeler's SLOCCount).

In my testing, memcpy_() is always faster than the standard memcpy() used on the Cortex-M0+ core. It is slower than the memcpy() used on the Cortex-M4, primarily because my implemention doesn't perform unaligned memory access.

This repo is designed to run on an STM32WLxx microcontroller. It uses the cycle counter of the Cortex-M4 microcontroller to profile the functions during testing. Note: the environment this repo was developed in is only important to know if you want to test the memcpy_() function like I did. If you just want to use it, follow the instructions in How to use this in an embedded project.

I also included testing using Greatest from https://github.com/silentbicycle/greatest. There's a number of tests that compare memcpy_() against the standard memcpy().