AI Software

SMD RISC-V SDK

Download SMD RISC-V SDK

Quickly and seamlessly develop, debug and fine-tune applications for Semidynamics RISC-V hardware with the SMD RISC-V SDK. It is a complete software development solution, including a compilation and debugging toolchain, emulators for functional testing of your applications and a highly optimized library of common routines, all integrated into a single Eclipse-based development environment.

SMD IDE

Development for RISC-V and Semidynamics hardware is simple with the SMD IDE. Compatible with Windows WSL or Linux environments, it is a complete solution integrated with all SDK features to enable users to code C and C++ applications with as little friction as possible. Immediately after installation, any user project can be compiled with the SMD toolchain and run with our SMD emulators.

Based on Eclipse, it includes all features required for software development optimization including debugging, a memory viewer, a disassembler and assembly viewer, code versioning, and much more. Complete compatibility with the universe of Eclipse plugins further allows users to customize and fine-tune their experience to their workflows.

  • Compatible with Linux and Windows Subsystem for Linux.

  • ASM/C/C++ software development environment.

  • Graphical debugging interface.

  • Integrated serial terminal and semihosting.

  • Integrated SMD emulators.

  • Integrated cross-compilation toolchain.

SMD Toolchain

SMD RISC-V SDK includes a comprehensive toolchain for compiling and debugging code targeting Semidynamics hardware. It is based on the GNU Compiler Collection GCC providing a common and familiar environment for users, with the full set of tools available for users including a debugger, linker and assembler. The compiler supports all the features expected of a modern compiler including code optimization and auto-vectorization. The toolchain is compatible with any IDE of your choice and is also seamlessly integrated in the SMD IDE for a complete development solution.

  • Compatible with Linux and Windows Subsystem for Linux.

  • C/C++ software development environment.

  • GCC based.

  • Auto-vectorization.

SMD Kernel Library

The SMD Kernel Library was created to most efficiently leverage the performance of the Semidynamics RISC-V hardware. It is a collection of functions operating on multi-dimensional data that have been maximally optimized for performance. With a particular focus on AI, where a model’s individual operations are called Kernels, a large number of crucial operations like Matrix multiplications, transpositions, activation functions, and more, have been optimized for Semidynamics hardware, enabling quick development of efficient AI applications.

  • Leverages Semidynamics Vector Unit.

  • Leverages Semidynamics Tensor unit.

  • Designed for multi-dimensional data.

  • Mathematical library.

  • Neural Network Library.

  • Data transformation.

  • Helper Functions.

SMD Emulators

The SMD RISC-V SDK includes the QEMU and SPIKE emulators that enables executing your code in your current Linux environment or even emulating baremetal executions. With complete support for RISC-V and SMD extensions, you can optimize and fine-tune your code quickly and ensure it is validated before ever needing silicon, derisking your development process.

  • User mode for bare-metal executions.

  • Semi-hosting for lightweight emulation without needing a virtual OS.

  • Up to 1 billion instructions per second for QEMU and up 100 million instructions per second for Spike.

  • Full support for all RISC-V extensions.

  • Full support for Semidynamics instructions.

SMD ONNX-Runtime

Download SMD ONNX-Runtime

By leveraging ONNX, Semidynamics hardware is ready to run thousands of models available in the ONNX format in common repositories like HuggingFace, or the ONNX Model Zoo. Using the SMD ONNX-Runtime, you’ll be able to integrate any of these models out of the box in Semidynamics hardware. The Microsoft led ONNX project defines a universal format for AI models and ONNX-RT enables any ONNX model to be executed in a diversity of hardware. Leveraging our Kernel Library, ONNX-RT is provided with support for Semidynamics hardware with Vector and Tensor units for the most efficient execution possible of AI applications.

AI Models

Model Zoo

Semidynamics is actively researching and developing the AI space and have therefore curated a set of AI models that we use as technology demonstrators. You can access these models in the ONNX format in our Model Zoo.

ONNX logo