The Unified Acceleration Foundation Aims to Boost ML, AI Workloads No Matter the Architecture
Cross-industry group including Arm, Intel, Qualcomm, and more looks to evolve oneAPI into an architecture-agnostic accelerator standard.
The Linux Foundation has announced the formation of the Unified Acceleration (UXL) Foundation, through which it hopes to lead a cross-industry group in the deliverance of an open standard for addressing accelerators for machine learning and more.
"The Unified Acceleration Foundation exemplifies the power of collaboration and the open source approach," claims Jim Zemlin, Linux Foundation executive director, of the new group. "By uniting leading technology companies and fostering an open ecosystem of cross-platform development, we will unlock new possibilities in performance and productivity for data-centric solutions."
The Foundation's founding members include Arm, Fujitsu, Google Cloud, Imagination Technologies, Intel, Qualcomm, and Samsung, and has been put together with one simple goal in mind: building on the success of the openAPI initiative to deliver an open programming model for accelerating workloads on CPU, GPU, FPGA, and dedicated accelerator hardware, regardless of architecture.
"The Foundation members aspire to build the largest open ecosystem for accelerated computing," claims Rod Burns, vice president for ecosystem at Codeplay Software and the UXL Foundation's steering committee chair. "The initial contributions to the foundation will bring an existing open standards based platform with open governance. Our ultimate aim is to foster a multi-architecture and multi-vendor programming platform for all accelerators,"
"As a founding member of the Unified Acceleration Foundation," adds Intel's Robert Cohn, who serves as editor of the oneAPI specification, "I believe that open source and standards are essential for creating a cross platform software stack for GPUs and other accelerators that will serve as the foundation for the next generation of computationally and data-intensive applications."
More information on the UXL Foundation is available on the official website; the oneAPI Specification on which the UXL standard will be based is available on GitHub under the Creative Commons Attribution 4.0 International license, with sample code published under the permissive MIT license.