UXL looks to standardise heterogeneous compute
Cette publication existe aussi en Français
Seven chip giants have come together to define standards for hetereogeneous compute that combines CPUs and GPUs.
ARM, Fujitsu, Google, Imagination Technologies, Intel, Qualcomm Technologies and Samsung are on the steering board of the Unified Acceleration (UXL) Foundation being hosted by the Linux Foundation.
“We are entering a new era of computer architectures, and it is highly heterogeneous,” said Rod Burns, VP ecosystem at Codeplay in Scotland, bought by Intel in 2022.
“This change brings a big opportunity to define and build a platform that brings performance portability. To achieve this goal we have announced UXL, hosted by the Linux Foundation’s Joint Development Foundation,” he said.
The aim is to define the specification for core libraries interfaces that implement fundamental operations for things like math and AI and have dependencies industry standards such as C++ and BLAS. The open source projects implement the specification and already have multi-vendor targets, with developers already writing code that is being run on AMD, ARM, Intel and NVIDIA processors from a single code base, and often a single binary.
“The UXL Foundation brings together founding Steering Members from across the industry and we also look forward to welcoming more organizations and individuals to join us on our mission. Together we can build a multi-architecture and multi-vendor programming model for all accelerators based on open source and standards.”
- The drivers behind Intel’s CodePlay software deal
- Codasip, Codeplay to combine RISC-V and SYCL
- Intel tapes out third generation Gaudi3 AI chip
This builds on developments such as Khronos, SYCL and oneAPI, which Intel and ARM see as an opportunity to dislodge Nvidia in AI applications using GPUs.
“My role at Codeplay for nearly seven years has involved helping to build a community around open standard programming for accelerators such as GPUs,” said Burns. “The primary focus was with SYCL™, a Khronos open standard, and since the first specification more than seven years ago a healthy ecosystem has evolved with multiple implementations, adoption from processor vendors and widespread use across the developer community. To complement this, over the past four years the oneAPI specification and projects have been developed.
“The oneAPI specification and implementation projects bring compatibility with SYCL using an open source compiler implementation based on LLVM Clang called DPC++. This makes it possible to enable multi-vendor targets whilst taking advantage of the power and performance portability offered by the SYCL programming model.”
- ARM hits the stock market with a bang with AI boost
- GPU cores with native HDR support in consumer AI chips
- How the data centre moved to custom chips
“A new open and collaborative approach to computing is required to provide the necessary acceleration in an efficient and performant manner, whether in the cloud or at the edge,” said Shreyas Derashri, Vice President of Compute Product Management at UK chip IP developer Imagination Technologies. “As a founding member of the Unified Acceleration Foundation, Imagination Technologies will help unite the technology ecosystem around the oneAPI spec and encourage its widespread adoption for compute and AI acceleration.”
The UXL foundation will build on this work with the oneAPI specification and open source projects being contributed to the foundation by Intel which will also collaborate with Khronos and ISO C++ who are developing standards that the UXL Foundation uses.
https://uxlfoundation.org/; www.codeplay.com
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
