How to use the Arm Cortex M55 Processor with the open source CMSIS library - stanlytw/CMSIS-NN GitHub Wiki
Get started with CMSIS on the Cortex-M55 processor
let’s look at vector multiplication using arm_mult_q31() from the CMSIS-DSP software library.
Cortex-M7
- Arm Compiler 6 uses 32-bit multiply instructions (SMMUL).
Cortex-M55 processor
- Uses vector multiply instructions and the Q registers (VQDMULH).
- Helium reuses the registers in the FPU as vector registers
- Each vector register is 128-bits wide. With Helium, the load, add, multiply, and store operations can be done using 128-bit values shown in the Q registers:

CMSIS Test Framework
- Can be used to do performance comparisons across the wide variety of functions provided by the CMSIS DSP library.
- This is a great way to get started doing performance analysis without the need to write any software.