White Papers
In recent years, the aerospace industry has sought to provide increased situational awareness for pilots and to improve operational efficiency of aircraft. This has resulted in new requirements for avionics display systems in the handling and display of more complex information. At the same time, there has been a disruptive technology change with the adoption of multi-core processors, which can provide significant benefits in terms of size, weight, and power (SWaP) but can also present challenges for RTCA DO-178C/EUROCAE ED-12C avionics software and RTCA DO-254/EUROCAE ED-80 avionics hardware safety certification.
To make the challenge complete, the relationship between original equipment manufacturers (OEMs), system integrators (Tier 1), and component providers (Tier 2) has changed drastically. OEMs are demanding more involvement in and control of the system design and more flexibility and configurability of the solution — in short, a more modular and more open system design approach, as reflected in the U.S. Department of Defense (DoD) Modular Open Systems Approach (MOSA) procurement directives.
This paper will discuss the development of a next-generation smart avionics platform, including:
• The use of an ARMv8 multi-core processor
• The use of hardware virtualization and a hypervisor for isolation and management of applications
• Runtime environments for safety-critical real-time applications
• The use of advanced Vulkan® SC-based GPU acceleration capabilities on modern graphics processing units (GPUs)
• The portability of software applications through use of ScioTeq’s MOSArt® and the ARINC 653 software architecture
In today’s increasingly automated world, decisions are made by algorithms interpreting the environment in which they operate. This is the plight of a car driving down a busy highway without human intervention. The car must understand its environment. It must be able to perceive road signs, lane markings, humans, animals, and other vehicles in its surroundings. The artificial intelligence system at the helm must decide how to steer the vehicle, when to accelerate, and when to apply the brakes. In the process of making these decisions there is a string of algorithms—often artificial neural networks—analyzing sensor data. How can we guarantee that the neural networks analyzing the input data coming from the sensors are making the right decisions? How do we know that when a neural network detects a human, there is indeed a human there, and can we know that it has not missed or misidentified other objects? Safe AI necessitates validating the expression of the decision-making system and ensuring the deterministic execution of the same system. Too often we forget about execution safety and rely on simulation to prove an AI system safe. As we will argue, simulation can solve part of the problem, but it cannot guarantee deterministic execution.
The introduction of advanced architectures has accelerated the transition to multi-core processors, bringing certification and validation challenges for avionics engineers who are building new solutions. The 13th Gen Intel® Core™ series of processors will greatly increase the advantages of a System-on-Chip (SoC) architecture that employs both “high performance” and “efficiency” cores. The two core types can run independently yet they share interfaces.
Interactive software systems with embedded safety features ensure the safety of the aircraft, passengers, and environment. Partnerships between cutting-edge software and hardware companies are also key to creating productive and more efficient certification processes for avionics engineers. CoreAVI has collaborated closely with Intel to port their best-in-class certifiable graphics and compute libraries to selected Intel processors. CoreAVI’s graphics and image processing libraries support the Khronos Group’s ratified Vulkan® SC™ standard, which allows avionics manufacturers to leverage Intel’s integrated graphics processing unit (GPU) to handle graphics and image processing, as well as neural networking applications. Platform applications that require the exposure of low-level GPU operations for critical operations will find both real time and deterministic support to meet the demands of safety critical requirements, such as DO-178C/ED-12C and DO-254/ED-80.
How can GPU acceleration benefit modern UAM/UAV applications? The explosion in machine learning and artificial intelligence research we see today is not due to fundamentally new computer science theories. Indeed, most of the math involved in these algorithms is decades old. The truly revolutionary thing in AI and ML today is the realization that we can accelerate the vast number of computations involved—which were once considered computationally intractable—using graphics processor units (GPUs). GPUs offer large arrays of computational cores. Provided that we write our algorithms in such a way that they can take advantage of the GPU cores to perform our computations, we can offload much of our algorithms to the GPU and leave the CPU free to continue running the system. This is true acceleration. This paper discusses the calculations involved in the use case of modeling the flight dynamics of a quadrotor, discusses which parts of the calculations are candidates for GPU acceleration, and demonstrates the importance of determinism to the execution time of these calculations.
Modern automotive applications are becoming more complex with the disruption of vehicle architecture, which requires heightened safety and managing of mixed-criticality applications. This development facilitates the merging of IVI and cluster, HUD, and mirror replacement displays. This paper, written jointly by CoreAVI, Arm and DiSTI, discusses moving the automotive digital cockpit to a safer and more reliable architecture while leveraging industry open standards and best-in-class partner collaborations, tools, and hardware to deliver a seamless workflow to developers. Using the latest technology innovations in hardware and software stacks, developers can rapidly create and iterate on a mixture of safe and standard embedded graphics and compute on new safety-critical GPUs. This approach enables faster time to market, reduces development costs and certification effort, and reduces risk, resulting in a safer user experience on and around the road.
Why is artificial intelligence (AI), and neural networks specifically, so popular these days? The math and the science behind these algorithms were developed decades ago, but it’s only in recent years that neural network-powered AI has taken off. So, what happened that enabled neural networks to succeed, where they fell short in the past?
Product developers often conduct trade studies before selecting components for their designs. One of the components that is often considered for these studies is a Graphics Processing Unit (GPU). GPUs have become a rather ubiquitous staple of most any electronic device or computer; however, the parameters that need to be collected to choose which GPU best fits a given application have grown increasingly complex. This paper addresses both discrete and embedded GPUs by comparing performance, cost, development environments, obsolescence, and certifiability. When all GPU variables and parameters are identified and quantified (or “costed”), a truly comprehensive design trade study can then be conducted by product/platform developers.
The Human Machine Interface (HMI) in embedded processing systems continues to become more complex. Aerospace, automotive, rail, and industrial control markets are pushing on all the edges of the technology “box”. Dis-plays are getting larger and of higher resolution, information content is becoming more complex and diverse, and images from sensors are now processed with sophisticated levels of Artificial Intelligence (AI) and Machine Learning (ML) algorithms. Further complicating these advances are pressures to make things more “open”, more portable, and safety-certifiable. This paper examines the evolving requirements for embedded processors that interact with pilots and opera-tors and suggests a solution that may bring harmony to these seemingly disparate requirements. A software architecture centered around the Vulkan® ecosystem will be described that holds the promise of more efficient processing for graphics, image processing, and autonomous decision making (through AI), while being bundled in an environment that is hardware agnostic and capable of being certified to the most stringent safety levels.
When topics of safety and artificial intelligence (AI) arise, the focus often rests on concerns around proving the intent of a neural network (NN). Although proving the intent of a network is an important problem to solve, it is not the only problem when it comes to the application of NN in real-time safety-critical systems (RTSCS). A key problem in this domain is proving that the platform execution is deterministic, meaning that the platform execution not only needs to provide results in a consistent and reliable fashion, but also needs to use a well-defined amount of memory. Existing AI platforms are built on technologies like Python that make use of techniques like runtime garbage collection. Time has shown that these existing platforms have not been a great fit for RTSCS. This paper explores how one would deterministically inference a NN, and how a standard like Khronos’ OpenVX™ provides a platform for AI that will be applicable for real-time safety-critical systems.
As electronics and software become more a part of everyday life, the reliability and predictability of these systems also increases in importance. Arm Ltd., the developer of the Arm processor, has licensed over 160 billion processor devices for phones, tablets, TVs, and even systems in airplanes and automobiles. This paper discusses Arm and their role in meeting the challenges related to safety certification within the graphics processing market. Safety certification challenges unique to GPU architectures will be addressed along with solution “eco-systems” related to solving the certification challenges for GPUs.
The new Arm Mali-G78AE GPU architecture provides a key innovation in its design: the ability to divide the GPU into partitions that effectively allow one Mali-G78AE GPU to be split into multiple independent mini Mali-G78AE GPUs. These “mini Mali-G78AE” GPU partitions are formed by combining GPU slices, where each partitioned GPU can have one or more slices assigned to it. The partitions provide all the functionality needed to operate as an independent GPU, including their own dedicated connection to the fabric to access memory, allowing each partition to function completely independent of the other partitions. This white paper describes how GPU Flexible Partitioning provides key benefits in the design of safety-critical systems.
This white paper describes methods for achieving diagnostic coverage on the Mali-G78AE GPU when used for graphics applications in Avionics and Automotive. While the discussion is specific to these industries, it addresses functional safety and applies to other applications requiring graphics with functional safety.
This paper examines the standards and guidance related to safety certification of object code generated by a compiler toolchain for both CPU and GPU targets. It is written in the context of two markets: avionics (DO-178C/ED-12C) and automotive (ISO 26262). (ISO 26262 is derived from the general IEC 61508, which is also used and derived for other markets with safety-critical applications, such as rail and nuclear). This paper will explain the motivation behind the guidance and identify approaches that may be used to address concerns. Understanding the guidance and constraints will facilitate the selection of an appropriate approach.
As automotive features and functions continuously evolve, so does the need for solutions for display and Advanced Driver Assistance Systems (ADAS). Displays have changed from analog to mixed analog/digital to completely digital, enabling other enhancements such as digital mirrors. ADAS is evolving from cruise control (maintenance of speed) to auto-pilot-like driver assistance, and beyond. This white paper discusses how these continuous advancements are increasing the need for GPUs to perform video processing and compute (data-level parallelism through many core SIMD engines) as well as advanced graphics.
As the Avionics industry standardizes on graphical processing Application Programming Interfaces (APIs), CoreAVI has been able to apply Research and Development (R&D) efforts to larger system level problems such as video latency. Video system latency is driven by many factors; however, through the use of clearly defined interfaces CoreAVI is able to offer a single product which can be configured to drive video system latency to a very low level.
This white paper examines the standards and guidelines for avionics and automotive safety-critical software and hardware to show how cost savings in commercial solutions can be achieved. We begin with a description of how software and hardware requirements are traditionally developed, followed by an examination of how guidelines for commercial solutions are developed ‘out-of-context’ of a typical safety application. Next, this paper will describe the process of selecting a solution and putting the solution into the safety application context. Finally, it details how safety certification is supported and describes examples of commercial solutions available and in use today.
This white paper provides a history of graphics and compute standards as well as graphics technology, and discusses the new Vulkan graphics and compute libraries available for specialist industries that have more stringent safety requirements such as aerospace, automotive and transportation.
This white paper details how existing safety critical DO-178C or ISO 26262 application software source code can effectively be rehosted on advancing hardware.
This white paper discusses six different mixed safety criticality scenarios for graphics rendering in embedded systems, their pros and cons, and use case considerations.
This white paper provides an introduction to the Vulkan API. It discusses Vulkan’s benefits and explains how it differs from OpenGL.
This white paper discusses the simultaneous failures that may occur due to common mode failures and how these can be mitigated through design diversity to meet the numerical safety requirements of the airplane.
This white paper details how a compositor works, the benefits and drawbacks of using different compositor solutions, and why using a compositor is conducive to safety certifiability to the most stringent levels for avionics, automotive, rail and other environments requiring safety critical operation.
This white paper examines the concerns and mitigations with using COTS Graphics Processors (CGPs or GPUs in general commercial terms) in safety critical applications requiring accelerated 2D and 3D safety rendering.
GPU architectures have vulnerabilities that could lead to unclassified applications accessing classified data, either maliciously or accidentally. This white paper describes the areas of vulnerability, consideration for multi-level security and how to support graphics applications requiring multi-level security
Modern multi-core processors and Real Time Operating Systems (RTOS) provide support for running multiple applications that improve performance, including graphics application performance. This white paper identifies the key architectures enabled by current multicore processors and RTOS to support multiple graphics applications and describes how OpenGL drivers can support these architectures.
Looking for low Size Weight and Power (SWaP) processing solutions without giving up high performance safety certifiable graphics? This joint white paper with AMD provides an introduction to solution worth considering.
If you are looking for a high performance graphics processor capable of driving multiple displays in an safety critical avionics system, then the AMD Radeon™ E8860 is a great choice. This joint white paper with AMD describes the benefits of the Radeon E8860 leading to its increased use on next generation commercial and military avionics applications, and why you may want to consider the Radeon E8860 too.
A new Safety Critical OpenGL® specification, OpenGL SC 2.0, was released by the Khronos Group April 2016. This paper describes how OpenGL SC 2.0 fits into the overall scheme of OpenGL specifications leading into a comparison to the earlier Safety Critical OpenGL specification, OpenGL SC 1.0.1, and concluding with an introduction to programmable shaders, now available to Safety Critical applications through OpenGL SC 2.0, enabling a higher degree of capability through new levels of performance and control.
The subject device is comprised of five (5) very large scale integrated circuits mounted on a high density multi-chip hybrid module. The part number of the hybrid module is 216T9NGBGA13FHG with a device description of ATI (now an AMD company) Mobility Radeon ™ 9000 M9-CSP64 Graphics Processor Unit, RoHS compliant. The module is an FR4 material Printed Circuit Board (PCB) mounted with an ATI designed GPU circuit in a plastic encapsulated Fine Ball Grid Array (FBGA) package. This is then mounted on the bottom (ball) side of the PCB and conformal coated…
COTS graphics processors (GPUs) have become popular components in mil-aero display systems with high performance graphics processing requirements. This article provides several GPU selection considerations that can impact the success of a display system design and delivery schedule as well as total life cycle systems management costs…
Case Studies
This case study highlights CoreAVI’s partnership with SYSGO and Ansys to create a joint safety critical railway solution addressing object detection and collision avoidance.
This case study highlights how CoreAVI’s partnership with Intel reduces customers’ mission computer system integration risks and speeds their time to deployment.
This case study discusses how CoreAVI’s Vulkan graphics and compute technology and COTS-D designs help enable HENSOLDT’s next generation airborne computer.
This case study demonstrates how CoreAVI’s Vulkan graphics and compute technology helps enable NASA to open up new possibilities for global supersonic air travel.
Newsletters
Product Briefs
GPMX002 XMC E9171 Graphics/Compute Processor
SBC3005 Intel Tiger Lake UP3 Single Board Computer
COTS-D Environmental Qualification
COTS-D factsheet
Platforms for Safety Critical Applications
VkCore® SC
VkCoreGL® SC1
VkCoreGL® SC2
ComputeCore™
VkCoreVX™ SC
ArgusCore SC™ 1
ArgusCore SC™ 2
EGL_EXT_Compositor: FACE-aligned Safety Critical Compositor
ArgusCore ES2/GL1.3
AMD Radeon E9171 GPU
AMD Radeon E8860 GPU
AMD Radeon E4690 GPU
AMD G Series SoC
Arm Mali-G78AE GPU
NXP i.MX 8 SoC
Intel Tiger Lake i7-1185GREC-RT SoC
S32V234 series of application processors
DecodeCore®
EncodeCore®
Hypercore™
TrueCore™ GPU health monitoring
CertCore178™: Avionics DO-178C/ED-12C Software Certification Data Packages
CertCore254™: Avionics DO-254/ED-80 GPU Certification Data Packages
CertCore26262™: Automotive ISO 26262 Certification Data Packages
JSF SEAL Software Certification Support
Videos
Joint Webinar: CoreAVI, DiSTI and NXP
Software reliability in today’s aerospace, automotive and industrial sectors is paramount. OEMs strive for complete UI flexibility while striking a balance between safety critical or functionally safe development features, practices, and costs. As entire systems require qualification to become a certifiable design, this webinar addresses the challenges and solutions for safety-critical graphics in tomorrow’s aerospace, automotive, and industrial use cases through the whole stack. From the top application layer to the middleware, drivers, operating systems, and down to the hardware, we will discuss the entire safety-critical systems stack and the best practices to accomplishing system certifiability.
Modern GPUs have many tricks up their sleeves, from hyper-parallelized computation, to efficient 3D rendering. Graphics APIs such as OpenGL, expose graphics functionality with a mindset that matches the GPUs of their era. This presentation will discuss the Vulkan API, a new modern graphics and compute API, which takes a radical shift from legacy APIs such as OpenGL and gives total control of the GPU to the application, allowing embedded graphics to reach the next level.
Core Avionics & Industrial Inc. and HENSOLDT Sensors GmbH have partnered to release the world’s first RTCA DO – 178 and EUROCAE ED-12C safely certifiable 4K video output hosted on HENSOLDT’s RTCA DO-254 and EUROCAE ED-80 safely certifiable Mission Computer with Curtiss-Wright’s COTS OpenVPX processor, I/O, and graphic module building blocks. This continues the long relationship between CoreAVI and HENSOLDT to provide innovative and cost-effective graphics and video processing solutions for safety critical applications such as synthetic vision systems (SVS).
This solution enables system integrators and end-users to leverage the high-resolution imagery provided by aircraft-installed sensors and available databases as well as large area displays to be installed in future aviation cockpits. Thus, the HENSOLDT Mission Computer with CoreAVI graphics drivers is already supporting the requirements of tomorrow’s avionic architectures.
CoreAVI brings flight displays to life powered by AMD G-Series Embedded processors and AMD Embedded Radeon™ Graphics.