Optimized Vector code is Useful.
A Scalable Verification Workflow is What Makes It Practical.

April 29, 2026

In our collaboration with TASKING around the AURIX™ TC4x PPU, the challenge was not only to generate efficient functions. The greater difficulty was ensuring consistent behavior under varying boundary conditions, such as:

  • different data sizes
  • different vector widths, including 256-bit and 512-bit
  • simulation and on-target execution
  • consistent checks for correctness, coverage, timing and numerical accuracy

That is why the project focused on a tool-assisted workflow that uses templates to combine function generation with matching test generation and verification.

The result was not just a set of optimized functions, but a structured ASPICE compatible framework for extending and validating PPU-ready implementations in a scalable way.

For teams working on vector-accelerated embedded software, the real bottleneck is often
not writing one fast function, but building a process that remains reliable as variants grow.

AI Workflow_Vectorization of ML Model
Cookie Consent with Real Cookie Banner