Analog Devices, Inc. simplifies embedded AI development with its latest CodeFusion Studio release, offering a new bring-your-own-model capability, unified configuration tools, and a Zephyr-based modular framework for runtime profiling. The upgraded open-source embedded development platform delivers advanced abstraction, AI integration, and automation tools to streamline the development and deployment of ADI’s processors and microcontrollers (MCUs).
CodeFusion Studio 2.0 is now the single entry point for development across all ADI hardware, supporting 27 products today, up from five in the last year, when first introduced in 2024.
Jason Griffin, ADI’s managing director, software and AI strategy, said the release of CodeFusion Studio 2.0 is a major leap forward in ADI’s developer-first journey, bringing an open extensible architecture across the company’s embedded ecosystem with innovation focused on simplicity, performance, and speed.
A major goal of CodeFusion Studio 2.0 is to help teams move faster from evaluation to deployment, Griffin said. “Everything from SDK [software development kit] setup and board configuration to example code deployment is automated or simplified.”
Griffin calls it a “complete evolution of how developers build on ADI technology,” by unifying embedded development, simplifying AI deployment, and providing performance visibility in one cohesive environment. “For developers and customers, this means faster design cycles, fewer barriers, and a shorter path from idea to production.”
A unified platform and streamlined workflow
CodeFusion Studio 2.0, based on Microsoft’s Visual Studio Code, features a built-in model compatibility checker, performance profiling tools, and optimization capabilities. The unified configuration tools reduce complexity across ADI’s hardware ecosystem.
The new Zephyr-based modular framework enables runtime AI/ML workload profiling, offering layer-by-layer analysis and integration with ADI’s heterogeneous platforms. This eliminates toolchain fragmentation, which simplifies ML deployment and reduces complexity, Griffin noted.
“One of the biggest challenges that developers face with multicore SoCs [system on chips] is juggling multiple IDEs [integrated development environments], toolchains, and debuggers,” Griffin explained. “Each core whether Arm, DSP [digital signal processor], or MPU [microprocessor] comes with its own setup and that fragmentation slows teams down.
“In CodeFusion Studio 2.0, that changes completely,” he added. “Everything now lives in a single unified workspace. You can configure, build, and debug every core from one environment, with shared memory maps, peripheral management, and consistent build dependencies. The result is a streamlined workflow that minimizes context switching and maximizes focus, so developers spend less time on setup and more time on system design and optimization.”
CodeFusion Studio System Planner also is updated to support multicore applications and expanded device compatibility. It now includes interactive memory allocation, improved peripherals setup, and streamlined pin assignment.
The growing complexity in managing cores, memory, and peripherals in embedded systems is becoming overwhelming, Griffin said. The system planner gives “developers a clear graphical view of the entire SoC, letting them visualize cores, assign peripherals, and define inter-core communication all in one workspace.”
In addition, with cross-core awareness, the environment validates shared resources automatically.
Another challenge is system optimization, which is addressed with multicore profiling tools, including the Zephyr AI profiler, system event viewer, and ELF file explorer.
“Understanding how the system behaves in real time, and finding where your performance can improve is where the Zephyr AI profiler comes in,” Griffin said. “It measures and optimizes AI workflows across ADI hardware from ultra-low-power edge devices to high-performance multicore systems. It supports frameworks like TensorFlow Lite Micro and TVM, profiling latency, memory and throughput in a consistent and streamlined way.”
Griffin said the system event viewer acts like a built-in logic analyzer, letting developers monitor events, set triggers, and stream data to see exactly how the system behaves. It’s invaluable for analyzing, synchronization, and timing across cores, he said.
The ELF file explorer provides a graphical map of memory and flash usage, helping teams make smarter optimized decisions.
CodeFusion Studio 2.0 also gives developers the ability to download SDKs, toolchains, and plugins on demand, with optional telemetry for diagnostic and multicore support.
Doubling down on AI
CodeFusion Studio 2.0 simplifies the development of AI-enabled embedded systems with support for complete end-to-end AI workflows. This enables developers to bring their own models and deploy them in ADI’s range of processors from low-power edge devices to high-performance DSPs.
“We’ve made the workflow dramatically easier,” Griffin said. “Developers can now import, convert, and deploy AI models directly to ADI hardware. No more stitching together separate tools. With the AI deployment tools, you can assign models to specific cores, verify compatibility, and profile performance before runtime, ensuring every model runs efficiently on the silicon right from the start.”
Easier debugging
CodeFusion Studio 2.0 also adds new integrated debugging features that bring real-time visibility across multicore and heterogeneous systems, enabling faster issue resolution, shorter debug cycles, and more intuitive troubleshooting in a unified debug experience.
One of the toughest parts of embedded development is debugging multicore systems, Griffin noted. “Each core runs its own firmware on its own schedule often with its own toolchain making full visibility a challenge.”
CodeFusion Studio 2.0 solves this problem, he said. “Our new unified debug experience gives developers real-time visibility across all cores—CPUs, DSPs, and MPUs—in one environment. You can trace interactions, inspect shared resources, and resolve issues faster without switching between tools.”
Developers spend more than 60% of their time doing debugging, Griffin said, and ADI wanted to address this challenge and reduce that time sink.
CodeFusion Studio 2.0 now includes core dump analysis and advanced GDB integration, which includes custom JSON and Python scripts for both Windows and Linux with multicore support.
A big advance is debugging with multicore GDP core dump analysis and RTOS awareness working together in one intelligent uniform experience, Griffin said.
“We’ve added core dump analysis, built around Zephyr RTOS, to automatically extract and visualize crash data; it helps pinpoint root causes quickly and confidently,” he continued. “And the new GDB toolbox provides advanced scripting performance, tracing and automation, making it the most capable debugging suite ADI has ever offered.”
The ultimate goal is to accelerate development and reduce risk for customers, which is what the unified workflows and automation provides, he added.
Future releases are expected to focus on deeper hardware-software integration, expanded runtime environments, and new capabilities, targeting growing developer requirements in physical AI.
CodeFusion Studio 2.0 is now available for download. Other resources include documentation and community support.