Codeplay Software Connecting AI to Silicon in Automotive

Charles Macfarlane is VP Marketing at Codeplay Software, leading marketing, sales and business development there for the past 4 years. Charles also engages with the leading global processor and semiconductor companies to provide leadership software solutions. Before Codeplay, Charles held positions as chip designer, applications engineer, product manager and marketing with major companies such as Broadcom and NXP. At these companies he was working with multimedia products for imaging, video and graphics in successful products for Nokia, Samsung, Sony Ericsson and Raspberry Pi. Charles holds an honour degree in Electronic Systems and Microprocessor Engineering from Glasgow University.

Codeplay is internationally recognized for expertise in heterogeneous systems and has many years of experience with open standards software such as OpenCL™, SYCL™ and Vulkan™ for complex processor architectures. Codeplay is enabling advanced vision processing and machine learning applications using ComputeAorta, an implementation of OpenCL for heterogeneous processing, and ComputeCpp™, a product based on the SYCL open standard for single-source programming using completely standard C++.

Codeplay is based in Edinburgh Scotland over 60 highly skilled software engineers. Codeplay has earned a reputation as one of the global leaders in compute processing systems.

Charles Macfarlane: Moore’s law is slowing down, CPUs have been stuck at 3GHz for many years. New heterogeneous processors can provide massive performance for artificial intelligence (AI), but only using specialist programming techniques. For example, most AI uses “graph programming” methods to enable individual AI operations to be combined reducing the bandwidth used and maximizing processing throughput.

AI and Machine Learning usage in almost all market segments has brought a hunger for specialist processors and therefore a demand for skilled engineering resources. Product development will therefore find the following barriers:
- Availability of familiar development frameworks and languages
- Availability of engineers with existing relevant skills
- Support during a product’s lifetime (especially in automotive)
- Avoiding lock-in implementations
- Avoiding legal and commercial issues
- Benefiting from mature and proven standards used in other markets
- Tracking the latest processor architectures
- Allowing application development and hardware processor solutions to evolve independently

Charles Macfarlane: Codeplay implements solutions based on established and widely adopted open standards. Codeplay works closely with The Khronos Group, an industry consortium focused on the creation of open standard, royalty-free application programming interfaces (APIs). Applications can now be developed using standard high-level C++ and deployed across heterogeneous processor systems without the need for specialized knowledge or skills for the underlying system. Our solutions also help connect AI to Silicon by using OpenCL. An example of this is our work on TensorFlow, Google’s popular AI framework. Codeplay’s SYCL implementation can be used to execute TensorFlow applications on any OpenCL enabled hardware.

Codeplay enables this by providing the following frameworks:
- ComputeAorta, an OpenCL open standards based solution for new specialized processors, making complex programmable devices easier to develop for by using well known programming standards, and
- ComputeCpp, a SYCL implementation enabling applications to be developed using standard high-level C++ and deployed across heterogeneous processor systems without the need for specialized knowledge or skills for the underlying system.

Charles Macfarlane: Reasons Codeplay is a leader and the first supplier considered for tough systems:
- Most Supported Platforms
- Working with the right customers driving the AI market
- Products already available and implemented
- Safest for product-ready implementation
- Fastest performance
- Based on widely adopted and understood standards
- Easiest to integrate

Charles Macfarlane: Automotive is experiencing huge growth in intelligent vision processing for Advanced Driver Assist Systems (ADAS) and ultimately into fully autonomous vehicles. Safety is a major driver for enabling cars with the latest AI innovations allowing cars to avoid accidents and save lives. Renesas is a leading global supplier of advanced system processors for cars and trucks, with their second-generation R-Car series enabling automotive firms to successfully implement a full range of emerging smart-car strategies. Codeplay’s open standards-based technology will be included in future cars so that Renesas’ R-Car solutions can interpret the surroundings and safely take control to avoid accidents or aid with driving functions.

Charles Macfarlane: Artificial Intelligence, in many forms, is already making an appearance in our lives, from voice devices to image recognition on our phones. We are incredibly early in the creation and adoption of smart devices, with greater intelligence in handheld devices, around the home, in the car and in industry, agriculture and medical – artificial intelligence can impact all parts of our lives in a very positive way.

So in the coming year we will see more specialised processor systems available for AI processing, rather than re-purposing existing solutions. Also the adoption of powerful AI frameworks such as TensorFlow will empower the programmer to build highly intelligent systems. These two advances in 2018 will bring greater-than-Moore’s law returns with huge steps forward in user experience. Codeplay has ensured ComputeAorta and ComputeCpp can address all market domains and processor types, providing open standards platforms for software developers. Codeplay’s extensive work with TensorFlow ensures programmers can benefit from the most popular processor platforms.