Codeplay Software is Safely Connecting AI to Silicon
07 May 2018
Artificial intelligence (AI) requires a very high level of software development expertise applying decades of research and mathematics, as well as making heavy demands of the best processor systems. With the evolution of semiconductor technology that delivers superior processor architectures and faster memory implementations, advanced AI can now be performed on devices scaling from powerful cloud compute systems performing tasks such as voice and video analysis, right down to embedded devices in homes and businesses used for the Internet-of-Things (IoT).
Today, many semiconductor and processor companies have their own specialized processor architecture tuned for complex AI functions. The latest AI applications are using neural networks to enable machine learning applications and these processors enable Neural Network (NN) operations to be performed with greater efficiency. However, as processor design tries to catch-up on AI needs, research continues to evolve extending the processing needs of the underlying hardware. So, you find a development environment where hardware and software must progress separately.
In software, an open standards layer is an interface specification developers base their solutions on. Similarly, processor developers create solutions with low-level software drivers that expose an open standards layer. The following benefits are then achieved:
Codeplay’s Major Collaborations
Since 2002, Codeplay has been working with global leaders to develop and deliver solutions into their next-generation systems. Codeplay is generally involved at the earliest stages of development to deliver tools that ease the development of applications into the most advanced processor designs.
Codeplay has demonstrated leadership and depth of knowledge through forward thinking and extensive experience. By strongly participating with open standards bodies and frequently delivering presentations at conferences, Codeplay’s reputation and ability to deliver is recognised globally within the leading technology companies. These companies come to Codeplay first when trying to solve the latest implementation challenges.
With Codeplay’s industry strength, experience and reputation, Codeplay is best positioned to take on the biggest integration challenges and enable AI everywhere.
Codeplay has already worked with the biggest and greatest companies located all around the world. However, achieving a reputation is not dependant on the company or location but achieved by demonstrating and delivering excellence.
One of the bigger challenges working with larger companies is aligning Codeplay’s technology within the corporation’s multiple organizations. While one group would like to enable Codeplay’s solution, other groups either believe they can do it quicker or cheaper, or other groups do not buy into the strategy. These challenges are often difficult to resolve in the short term and take many months of follow-up to convince and align minds.
All companies have incredibly smart engineers and one of the challenges is to ensure they are aware of the benefits that can be achieved by using Codeplay's technology. Understanding the hierarchy within companies and how many groups relate to each other will help focus relevant information flow without excessive sales push.
Face-to-face is great, it totally breaks down barriers and avoids misinterpretation. Whiteboard brainstorming, and sketching thoughts always helps to communicate any concerns.
Every company wants Codeplay's technology delivered tomorrow but it can take well over 6 months to get a project kicked-off. Engineering level discussions are great and progressive but it is time efficient to engage sourcing departments in parallel to get all the supplier details sorted out e.g. master services agreements, supplier acceptance and commercial terms.
AI Market Segments
AI is everywhere today where some applications are more visible to consumers than others but almost all are starting to embrace systems that Codeplay can enable:
Smart home: from cameras identifying object/people, to thermostats and security devices
Audio solutions: intelligent speakers in the home and smartphone
Industry: intelligent manufacturing and handling systems
Medical: analysis and diagnosis used to propose bespoke treatments
Cloud servers: with massive quantities of data being processed with the deepest algorithms
Finance: for interpreting, predicting and reacting to market trends
Automotive using AI is now ripe for disruption. Vision processing and machine learning systems in cars allow decisions to be taken from the driver. ADAS (Advanced Driver Assistance Systems) features are already in most high-tier cars providing features such as adaptive cruise control, self parking, lane departure control and collision avoidance. These features and many more advanced functions will continue to trickle into mid- and low-tier cars providing enhanced safety to drivers and road users. All car manufacturers and component manufacturers are scrambling to produce solutions embracing AI research and vision processing to benefit the high-value and high-reward ADAS industry.Achieving AI Everywhere
Open standards are the easiest way to enable the integration of AI into any product. By bringing relevant industry leaders together (processor developers, semiconductor manufacturers and software developers) the whole industry can agree the interface layers, known as APIs (Application Program Interface), which means everyone is developing in a common and compatible way.
Codeplay has achieved a leading reputation for enabling the toughest and most capable processors with open standards-based solutions. By using widely agreed and understood open standards, and revolutionizing the PC, mobile and gaming markets, these implementations can now be used to bring structure and efficiency to AI industries.
There will always be first movers who bring proprietary solutions to market first, who create their own infrastructure and make it difficult to transition to other platforms. Generally, these solutions are non-optimal, lack choice, lack flexibility, are costly and leave little space for differentiation. While these solutions are great for bringing innovative products to the market, there are so many AI innovations and innovators wanting to bring their valuable creations to market across different hardware systems.
As development costs increase, car manufacturers are already finding that software development already accounts for over 50% of their costs. But this software cost is escalating and OEMs are seeking a route to control these costs while integrating leading ADAS features. Additionally, in order for these systems to continue to be safe, they are evolving to adopt some of the same features seen in smartphones, such as over the air software updates. This in turn requires long term after-sales support, perhaps between 10 and 20 years.
Renesas, a Japanese semiconductor manufacturer and major supplier of automotive processors, is enabling the next generation of ADAS features and autonomous car solutions. They quickly aligned with Codeplay’s vision for building a software infrastructure based on open standards to simplify and accelerate AI deployment. This is a major engagement for Codeplay, Renesas are endorsing Codeplay’s strategy and vision. It also sends out a major statement within the automotive industry that enabling AI and vision processing using open standards-based software is the right way to go.
Codeplay's tools enable companies developing applications to focus on their ADAS innovation, writing software using a familiar environment. Codeplay's open standard stack allows these applications to be run on a range of Renesas’ R-Car processors.
The collaboration with Renesas is not exclusive and Codeplay will extend the solution with other processor suppliers, in automotive and all other AI markets.
There are 6 levels of autonomy, commonly referred to SAE levels:
As of today, only one or two high-end car models claim to be at Level 3 Conditional Automation (fusing high quality maps, radar and sensors), but most cars now sold have a minimum of Level 1 with most mid-tier and high end containing Level 2 solutions. Therefore, the route to Level 5 Full Automation will take a long time, with some people saying first introductions will happen around 2025 to 2030, and others believe it will never happen.
There are so many steps before we get there, with combinations of implementations building up over time, each step releasing control from the driver and adding extra safety and automation. Level 4 is forecast to be introduced in production cars some time after 2020 and can allow the car to be autonomous with certain limitations e.g. only on approved/verified roads, accepted town centres, acceptable weather conditions & platooning lorries where the leading lorry has driver, and the followers are driverless. This type of autonomous car already takes us far beyond today’s position and is believed to be sufficient to change the car ownership business model (the future of car ownership is another subject).
We are already seeing an array of sensors in prototype cars i.e. radar, LiDAR, vision and ultrasonic. These core sensors can be supplemented with other sources of data e.g. GPS, wheel speeds, steering wheel position. Also, cameras monitoring the driver to ensure awareness of a perceived risk e.g. the car approaching an obstacle while the driver is sleeping or distracted. All the sources of data need fusing, interpreting and taking the most appropriate manoeuvre.
Codeplay’s solution is ideally structured for all applications, from the reduced feature peripheral imaging devices detecting specific features, the array of intermediate implementations, through to a high-end sensor-fusion with advanced intelligence. By enabling the developers of advanced innovation, Codeplay’s solutions will provide support for applications with processors that have not even yet been developed. The use of open standards provides the glue between applications and processors.
Smart homes, mobile phones, manufacturing and medicine are starting to benefit from AI. This will evolve significantly in the coming years, with the computing power moving from cloud servers to embedded devices.
For Codeplay, automotive is the biggest challenge and the biggest reward. Saving lives and reducing accidents on the roads is achievable with AI, from ADAS to autonomous cars in some form. Therefore Codeplay is substantially evolving the safety critical certification of ComputeSuite.
One open standard Codeplay is supporting, which is highly relevant to enabling AI everywhere, is SYCL™ from the Khronos Group, an industry consortium focused on the creation of open standard, royalty-free application programming interfaces (APIs). A SYCL enabled processor system provides application developers with a familiar application programming interface and enables many familiar AI and machine learning solutions.
SYCL has been gathering momentum over the last few years with Codeplay leading this push. Codeplay offers a free implementation called ComputeCpp™ Community Edition, which has thousands of downloads and received excellent feedback. Codeplay is also driving an eco-system bringing developers, news, releases and updates into one place – see http://sycl.tech. In the last few months many companies are understanding its importance and SYCL is therefore getting much more interest and attention.
The moment for industry adoption of SYCL as the platform getting AI everywhere is here and 2018 will certainly be an interesting year for Codeplay and SYCL.
Today, many semiconductor and processor companies have their own specialized processor architecture tuned for complex AI functions. The latest AI applications are using neural networks to enable machine learning applications and these processors enable Neural Network (NN) operations to be performed with greater efficiency. However, as processor design tries to catch-up on AI needs, research continues to evolve extending the processing needs of the underlying hardware. So, you find a development environment where hardware and software must progress separately.
In software, an open standards layer is an interface specification developers base their solutions on. Similarly, processor developers create solutions with low-level software drivers that expose an open standards layer. The following benefits are then achieved:
- Application developers can develop software without being bound to a specific hardware solution
- Processor developers can provide hardware capable of supporting existing applications as well as enabling a larger number of application developers
- Wider skills availability, where engineers can work with development tools they are familiar with
- Accelerated time to market and long-term maintenance reduced
- Industry wide acceptance of open standards layers, with all major companies contributing to their definitions
Codeplay’s Major Collaborations
Since 2002, Codeplay has been working with global leaders to develop and deliver solutions into their next-generation systems. Codeplay is generally involved at the earliest stages of development to deliver tools that ease the development of applications into the most advanced processor designs.Codeplay has demonstrated leadership and depth of knowledge through forward thinking and extensive experience. By strongly participating with open standards bodies and frequently delivering presentations at conferences, Codeplay’s reputation and ability to deliver is recognised globally within the leading technology companies. These companies come to Codeplay first when trying to solve the latest implementation challenges.
With Codeplay’s industry strength, experience and reputation, Codeplay is best positioned to take on the biggest integration challenges and enable AI everywhere.
Codeplay has already worked with the biggest and greatest companies located all around the world. However, achieving a reputation is not dependant on the company or location but achieved by demonstrating and delivering excellence.
One of the bigger challenges working with larger companies is aligning Codeplay’s technology within the corporation’s multiple organizations. While one group would like to enable Codeplay’s solution, other groups either believe they can do it quicker or cheaper, or other groups do not buy into the strategy. These challenges are often difficult to resolve in the short term and take many months of follow-up to convince and align minds.
All companies have incredibly smart engineers and one of the challenges is to ensure they are aware of the benefits that can be achieved by using Codeplay's technology. Understanding the hierarchy within companies and how many groups relate to each other will help focus relevant information flow without excessive sales push.
Face-to-face is great, it totally breaks down barriers and avoids misinterpretation. Whiteboard brainstorming, and sketching thoughts always helps to communicate any concerns.
Every company wants Codeplay's technology delivered tomorrow but it can take well over 6 months to get a project kicked-off. Engineering level discussions are great and progressive but it is time efficient to engage sourcing departments in parallel to get all the supplier details sorted out e.g. master services agreements, supplier acceptance and commercial terms.
AI Market Segments
AI is everywhere today where some applications are more visible to consumers than others but almost all are starting to embrace systems that Codeplay can enable:Smart home: from cameras identifying object/people, to thermostats and security devices
Audio solutions: intelligent speakers in the home and smartphone
Industry: intelligent manufacturing and handling systems
Medical: analysis and diagnosis used to propose bespoke treatments
Cloud servers: with massive quantities of data being processed with the deepest algorithms
Finance: for interpreting, predicting and reacting to market trends
Automotive using AI is now ripe for disruption. Vision processing and machine learning systems in cars allow decisions to be taken from the driver. ADAS (Advanced Driver Assistance Systems) features are already in most high-tier cars providing features such as adaptive cruise control, self parking, lane departure control and collision avoidance. These features and many more advanced functions will continue to trickle into mid- and low-tier cars providing enhanced safety to drivers and road users. All car manufacturers and component manufacturers are scrambling to produce solutions embracing AI research and vision processing to benefit the high-value and high-reward ADAS industry.Achieving AI Everywhere
Open standards are the easiest way to enable the integration of AI into any product. By bringing relevant industry leaders together (processor developers, semiconductor manufacturers and software developers) the whole industry can agree the interface layers, known as APIs (Application Program Interface), which means everyone is developing in a common and compatible way.
Codeplay has achieved a leading reputation for enabling the toughest and most capable processors with open standards-based solutions. By using widely agreed and understood open standards, and revolutionizing the PC, mobile and gaming markets, these implementations can now be used to bring structure and efficiency to AI industries.
There will always be first movers who bring proprietary solutions to market first, who create their own infrastructure and make it difficult to transition to other platforms. Generally, these solutions are non-optimal, lack choice, lack flexibility, are costly and leave little space for differentiation. While these solutions are great for bringing innovative products to the market, there are so many AI innovations and innovators wanting to bring their valuable creations to market across different hardware systems.
Enabling Safe AI in Automotive
Safe AI in automotive will be one of the great uses of the technology. ADAS is already widely deployed into many cars but the level of functionality is still low, costs are high and still so much innovation is to be included.As development costs increase, car manufacturers are already finding that software development already accounts for over 50% of their costs. But this software cost is escalating and OEMs are seeking a route to control these costs while integrating leading ADAS features. Additionally, in order for these systems to continue to be safe, they are evolving to adopt some of the same features seen in smartphones, such as over the air software updates. This in turn requires long term after-sales support, perhaps between 10 and 20 years.
Renesas, a Japanese semiconductor manufacturer and major supplier of automotive processors, is enabling the next generation of ADAS features and autonomous car solutions. They quickly aligned with Codeplay’s vision for building a software infrastructure based on open standards to simplify and accelerate AI deployment. This is a major engagement for Codeplay, Renesas are endorsing Codeplay’s strategy and vision. It also sends out a major statement within the automotive industry that enabling AI and vision processing using open standards-based software is the right way to go.
Codeplay's tools enable companies developing applications to focus on their ADAS innovation, writing software using a familiar environment. Codeplay's open standard stack allows these applications to be run on a range of Renesas’ R-Car processors.
The collaboration with Renesas is not exclusive and Codeplay will extend the solution with other processor suppliers, in automotive and all other AI markets.
There are 6 levels of autonomy, commonly referred to SAE levels:
As of today, only one or two high-end car models claim to be at Level 3 Conditional Automation (fusing high quality maps, radar and sensors), but most cars now sold have a minimum of Level 1 with most mid-tier and high end containing Level 2 solutions. Therefore, the route to Level 5 Full Automation will take a long time, with some people saying first introductions will happen around 2025 to 2030, and others believe it will never happen.
There are so many steps before we get there, with combinations of implementations building up over time, each step releasing control from the driver and adding extra safety and automation. Level 4 is forecast to be introduced in production cars some time after 2020 and can allow the car to be autonomous with certain limitations e.g. only on approved/verified roads, accepted town centres, acceptable weather conditions & platooning lorries where the leading lorry has driver, and the followers are driverless. This type of autonomous car already takes us far beyond today’s position and is believed to be sufficient to change the car ownership business model (the future of car ownership is another subject).
We are already seeing an array of sensors in prototype cars i.e. radar, LiDAR, vision and ultrasonic. These core sensors can be supplemented with other sources of data e.g. GPS, wheel speeds, steering wheel position. Also, cameras monitoring the driver to ensure awareness of a perceived risk e.g. the car approaching an obstacle while the driver is sleeping or distracted. All the sources of data need fusing, interpreting and taking the most appropriate manoeuvre.
Codeplay’s solution is ideally structured for all applications, from the reduced feature peripheral imaging devices detecting specific features, the array of intermediate implementations, through to a high-end sensor-fusion with advanced intelligence. By enabling the developers of advanced innovation, Codeplay’s solutions will provide support for applications with processors that have not even yet been developed. The use of open standards provides the glue between applications and processors.
Codeplay’s Ambition
AI is all around us yet still at the start of the growth curve, having passed the hype stage. AI has only just touched us so far, yet consumers are already experiencing it positively. There is however excitement in a range of markets that believe AI could do so much more and companies are making it part of their strategy or roadmap.Smart homes, mobile phones, manufacturing and medicine are starting to benefit from AI. This will evolve significantly in the coming years, with the computing power moving from cloud servers to embedded devices.
For Codeplay, automotive is the biggest challenge and the biggest reward. Saving lives and reducing accidents on the roads is achievable with AI, from ADAS to autonomous cars in some form. Therefore Codeplay is substantially evolving the safety critical certification of ComputeSuite.
One open standard Codeplay is supporting, which is highly relevant to enabling AI everywhere, is SYCL™ from the Khronos Group, an industry consortium focused on the creation of open standard, royalty-free application programming interfaces (APIs). A SYCL enabled processor system provides application developers with a familiar application programming interface and enables many familiar AI and machine learning solutions.
SYCL has been gathering momentum over the last few years with Codeplay leading this push. Codeplay offers a free implementation called ComputeCpp™ Community Edition, which has thousands of downloads and received excellent feedback. Codeplay is also driving an eco-system bringing developers, news, releases and updates into one place – see http://sycl.tech. In the last few months many companies are understanding its importance and SYCL is therefore getting much more interest and attention.
The moment for industry adoption of SYCL as the platform getting AI everywhere is here and 2018 will certainly be an interesting year for Codeplay and SYCL.
Codeplay Software Ltd has published this article only as an opinion piece. Although every
effort has been made to ensure the information contained in this post is accurate and
reliable, Codeplay cannot and does not guarantee the accuracy, validity or
completeness of this information. The information contained within this blog
is provided "as is" without any representations or warranties, expressed or implied.
Codeplay Sofware Ltd makes no representations or warranties in relation to
the information in this post.
