2017 - The Year for Connecting AI to Silicon

Posted on December 19, 2017 by Charles Macfarlane.

2017 has been an exciting year seeing artificial intelligence emerging within most market segments and Codeplay's tools are available for developers to create intelligent products from self-driving cars to very low-power AI devices using small development boards. In addition, new processor architectures are being developed to run these AI solutions, making compute more available on these devices, reducing the need for big data transfer and extensive cloud servers.

Developers are looking to rapidly create AI solutions for these systems and Codeplay’s products are essential in helping those developers to use a familiar programming environment to build high performance AI software.

During 2017 Codeplay continued adapting Google’s machine learning framework TensorFlow™ to work with Khronos’ SYCL™, an open standard solution providing a open standard C++ interface for application developers. This work has made it possible to run and accelerate TensorFlow applications on a range of OpenCL™ hardware including Arm Mali GPUs. By using OpenCL and ComputeCpp™, Codeplay’s implementation of SYCL, we have been able to bring the benefits of speed of integration and performance to TensorFlow. The benchmark results from HiSilicon’s Kirin 960 with TensorFlow were presented at TechCon, Arm’s annual technical conference .

Automotive vision solutions continue to gain momentum with substantial investment into all implementation options. As more and more complex ADAS units are integrated into cars and bring safety technology to all types, Codeplay was proud to announce a partnership with Renesas, the leading automotive processor supplier, to provide ADAS application developers with a platform that will benefit from safety critical implementation of OpenCL and SYCL.

Codeplay’s work also captured the attention in many competitions this year, but most importantly winning CW’s Discovering Startup competition award. This highlights the forward-looking approach used to deliver new AI solutions.

At home in Edinburgh, Codeplay’s team of talented engineers expands while supporting universities via their highly-respected intern program, with a record 15 intern records. The ecosystem around SYCL grew in 2017 with enhancements to SyclParallelSTL, a high level framework offering a simple interface for parallel software development, and also the use of SYCL to accelerate Eigen, one of the most popular linear algebra frameworks. ComputeCpp, Codeplay’s implementation of SYCL, now offers Windows support, support for ARM hardware, experimental support for PTX to enable NVIDIA hardware and developers can follow a new beginners guide to SYCL alongside a range of sample code and projects.

Codeplay continues to support open standards, especially chairing Khronos’ SYCL, which saw the ratification of SYCL 1.2.1 , and chair of Khronos’ Safety Critical Advisory Panel.

So to 2018, beyond investment activities, Codeplay will focus on two main areas:

  • Bringing intelligence to mobile and IoT, delivering high performance and maintainable implementations
  • Automotive platforms, growing the developer ecosystem via open standards and safety critical robustness

Exciting times at the Edinburgh base, experiencing the pull of the evolving AI markets and the brilliant Codeplay team with products ready for growth.