ADI's CodeFusion Studio 2.0 speeds up embedded AI development
Analog Devices Inc. (ADI) has updated CodeFusion Studio with a bring-your-own-model workflow, unified configuration tools, and a Zephyr-based framework for runtime profiling. The open-source platform is now the single entry point for development across ADI hardware, growing from five to 27 supported products since 2024. The focus: simplify setup, improve visibility, and shorten the path from idea to production.
"Everything from SDK setup and board configuration to example code deployment is automated or simplified," said Jason Griffin, ADI's managing director of software and AI strategy. He calls it a complete evolution in how teams build on ADI technology by unifying embedded development, streamlining AI deployment, and providing performance visibility in one place.
Why it matters for dev teams
- One workspace for multicore SoCs-configure, build, and debug across CPUs, DSPs, and MPUs without juggling multiple IDEs.
- Consistent tooling and dependencies reduce context switching and setup time.
- End-to-end AI workflow with model import, conversion, validation, and pre-runtime profiling.
- Deeper insight into latency, memory, and throughput before you flash.
Key upgrades at a glance
- Bring-your-own-model (BYOM): Import, convert, validate, and deploy models to ADI processors with a built-in compatibility checker and pre-runtime performance estimates.
- Unified configuration: Shared memory maps, peripheral management, and consistent build dependencies across heterogeneous platforms.
- Zephyr-based profiling: Layer-by-layer AI/ML analysis with a modular framework tied into ADI's heterogeneous platforms. Supports TensorFlow Lite Micro and TVM.
- System Planner updates: Interactive memory allocation, improved peripheral setup, streamlined pin assignment, and cross-core resource validation.
- Multicore profiling tools: Zephyr AI profiler for latency, memory, and throughput; system event viewer with triggers and streaming; ELF file explorer for memory and flash mapping.
- On-demand components: Download SDKs, toolchains, and plugins as needed with optional telemetry.
- Built on VS Code: A familiar foundation with extensibility and clean integration.
Unified multicore workflow
"One of the biggest challenges with multicore SoCs is juggling multiple IDEs, toolchains, and debuggers," Griffin said. CodeFusion Studio 2.0 moves it all into one workspace. Configure, build, and debug every core with shared memory maps and consistent dependencies. Less setup friction, more time for system design and optimization.
Debugging that scales with heterogeneous systems
CodeFusion Studio 2.0 adds a unified debug experience with real-time visibility across CPUs, DSPs, and MPUs. Trace interactions, inspect shared resources, and resolve issues without switching tools. It's built to reduce the 60% of time developers often spend debugging.
The release also includes core dump analysis (with Zephyr RTOS awareness) and advanced GDB integration. You get custom JSON and Python scripting on Windows and Linux, plus tracing and automation through a new GDB toolbox.
AI deployment made practical
Developers can assign models to specific cores, verify compatibility, and profile performance before runtime. "You can import, convert, and deploy AI models directly to ADI hardware," Griffin said. The workflow covers everything from small edge devices to high-performance multicore systems.
Practical first steps
- Install or upgrade to CodeFusion Studio 2.0 and connect your target board.
- Use System Planner to map cores, peripherals, memory regions, and shared resources.
- Import a small TensorFlow Lite Micro model, run the compatibility checker, and assign it to the right core.
- Run the Zephyr AI profiler to validate latency, memory, and throughput before flashing.
- Set triggers in the system event viewer and review the ELF map for memory pressure.
- Enable optional telemetry (if allowed) for diagnostics and multicore support.
- Test the unified debug flow with RTOS awareness, then capture a synthetic crash and review core dump analysis.
Availability and what's next
CodeFusion Studio 2.0 is available now with documentation and community support. ADI plans deeper hardware-software integration, expanded runtime environments, and features aligned with growing needs in physical AI.
Resources
- Visual Studio Code (foundation for CodeFusion Studio)
- Zephyr Project (RTOS and tooling ecosystem)
- AI courses by job for teams building embedded AI workflows
Your membership also unlocks:
