The oneAPI DevSummit is an event focused on the oneAPI specification—a cross-industry, open programming model designed by Intel to support diverse hardware architectures.
Held globally and often hosted by organizations like the UXL Foundation, the DevSummit series brings together developers, researchers, and industry leaders to explore the practical applications of oneAPI in areas such as AI, high-performance computing (HPC), edge computing, and beyond.
At the recent oneAPI DevSummit hosted by the UXL Foundation, Intel® Liftoff member, Roofline AI took center stage to highlight its innovative approach to enhancing AI and high-performance HPC performance.
Their pitch addressed a central need in the AI and HPC ecosystem: efficient and adaptable AI compiler support that can seamlessly integrate with various devices, a goal realized through Roofline AI’s integration with the UXL framework.
AI compilers play a critical role in bridging AI models and the hardware that executes them. In their pitch, Roofline AI’s team emphasized that by using the open-source Multi-Level Intermediate Representation (MLIR), they’ve created a robust compiler that supports end-to-end model execution for the UXL ecosystem. This architecture gives developers the ability to map and execute AI models across different devices with unparalleled efficiency and flexibility.
It’s a definite step forward for hardware-agnostic AI processing, particularly suited to industries with diverse hardware demands.
At the heart of their solution is a lightweight runtime built on the Level Zero API, which calls kernels and manages memory with high efficiency.
Roofline AI’s runtime not only optimizes execution but also ensures compatibility with a wide range of hardware that supports Level Zero, including Intel GPUs. This compatibility allows their software to control devices out-of-the-box, minimizing configuration needs, and broadening the hardware options available to developers.
During the presentation, Roofline AI demonstrated how its compiler translates machine learning models from popular frameworks like PyTorch and TensorFlow into SPIR-V code; a specialized language for executing parallel compute tasks.
The result is a streamlined process that enables rapid, optimized AI model deployment across multiple platforms, making it easier for developers to achieve top performance without custom configurations for each hardware type.
Roofline AI’s commitment to enhancing compiler technology exemplifies the potential of oneAPI in supporting next-generation AI. With unified support for various devices and streamlined integration with the UXL ecosystem, Roofline AI is not only improving AI deployment but setting a new standard for AI scalability and efficiency.
As they continue to push the boundaries of AI compiler technology, Roofline AI is positioning itself as a key player in the future of scalable, high-performance AI applications.
For more information on the Intel® Liftoff for Startups program, visit www.intel.com/content/www/us/en/developer/tools/oneapi/liftoff.html
As a Software Tools Ecosystem Specialist at Intel, I’ve had the privilege of working on the dynamic GenAI initiative. My focus is on driving engagement with the software developer audience. I'm a proud team member of the Intel® Liftoff for Startups and the Developer Engagement, Relations, and Studio team.