Is it possible for Simulink to support m7 cortex for NXP and requires an third party tool?

As a automotive professional, I trust MATLAB & Simulink. But when it comes to third party tools, I lack trust as Mathworks cannot take responsibility for any issue caused by the third party tools.

I want to generate application code for NXP S32K3XX controller which is ARM cortex M7 tuned for double precision datatype with 6 stage pipeline. But generic matlab ert. allows only till M4(3-stage pipeline). I'll be using 32 bit single precision. Even though I'm not going to use 64bit double precision datatype, I hope there will be some advantages M7 compared to M4. I would like to know what kind of difference in the performance while using 32 bit single datatype in M4 & m7

If the performance difference is drastic, If Mathworks have any Idea to bring M7 for ert. codegen in future?

Answer 1

[Ritam]

When using 32‑bit single precision, ARM Cortex‑M7 generally offers a significant performance boost over Cortex‑M4. The M4 core delivers about 1.25 DMIPS/MHz with a 3‑stage pipeline and optional single‑precision FPU, while the M7 achieves roughly 2.14 DMIPS/MHz thanks to its 6‑stage dual‑issue pipeline, branch prediction, and improved memory system with I/D‑cache and TCM. In real‑world control and DSP workloads, this translates to about 1.5 to 2.5 times faster execution at the same clock frequency, provided caches and TCM are enabled and code/data are placed correctly. If these features are not configured, the advantage can shrink considerably. In short, even without double precision, M7’s architecture and memory optimizations make it a much stronger choice for single‑precision applications.