New Release: emmtrix ONNX-to-C Code Generator (emx-onnx-cgen) v1.0.0

March 10, 2026

With version 1.0.0, emx-onnx-cgen reaches an important milestone: 100% operator coverage for ONNX Opset 25.
All operators defined in the current ONNX specification are now supported by the compiler pipeline (data type coverage is still being expanded).

We are proud to reach this milestone.

What is emx-onnx-cgen?
The emmtrix ONNX-to-C Code Generator (emx-onnx-cgen) translates ONNX models into portable, standards-compliant C code. The generated code provides a clean and analyzable baseline for further compiler optimizations and deployment in embedded, safety-critical, and performance-sensitive environments.

What ist new in v1.0.0?

  • Full ONNX Opset 25 operator coverage across neural network, ML, signal processing, and training operators
  • Major additions including operators such as TreeEnsemble, RoiAlign, ScatterElements, MaxUnpool, STFT, MelWeightMatrix, Adam, and more
  • Expanded quantization support including additional QLinear operators and improved zero-point handling
  • Stronger verification framework with per-node accuracy reporting and deterministic validation
  • Improved numerical robustness with wider accumulators and more stable kernels
  • Better support for sequences, optionals, and control-flow operators such as If and Loop
  • Improvements across CLI, CI, and documentation

This release represents a significant step toward making ONNX-to-C compilation robust enough for real-world deployment.

AI Workflow_Vectorization of ML Model
Cookie Consent with Real Cookie Banner