MENU

GPUs – a great fit for the challenges of ADAS

Feature articles |
By Rich Pell


These systems rely on higher-quality data from an increasing number of discrete sensors, such as Light Detection and Ranging (LiDAR), which measures the distance to a target with a pulsed laser light, Radio Detection and Ranging (RADAR), which is similar to LiDAR but uses radio waves instead of a laser, and infrared (IR) cameras systems. These all enable ADAS to better interpret the environment and improve its ability to help the driver.

Introducing the GPU
Within a car’s central electronic control unit (ECU), you will find one or more large silicon devices that contain a multi-core central processing unit (CPU), a graphics processing unit (GPU), a memory subsystem that feeds the sensor data for real-time processing, and a range of other cores, such as I/O, peripheral connection, dedicated video and DSP.

The GPU’s highly parallel, throughput-oriented nature makes it a great fit for the challenges of ADAS. It is essentially a turbo-charged multiply, accumulate engine which is the basis of neural network-type algorithms. Therefore, it’s no wonder that many of today’s leading ADAS vendors are starting to exploit the ability and performance of embedded GPUs in order to make the generational leaps in capabilities and performance required over what is available in cars today.

Historically, the GPU plays a more traditional role in a car’s technical makeup – namely driving the displays. Consumers increasingly expect that their new car offers the same cutting-edge technology they are used to from their smartphone or tablet. Therefore, we are seeing a move to digital dashboards with multiple high-resolution screens. 1080p is now common in mid-range cars and increasingly 4K screens are specified for luxury and executive cars.


These larger, higher-resolution and more responsive screens are enabling drivers and passengers to interact with their car in more natural and intuitive ways. It is, therefore, a necessity – not merely a nice to have – for car OEMs to move from smaller CPUs or microcontrollers to more powerful GPUs to drive those screens, and given this, embedded GPUs have made obvious inroads in this area.

Why GPUs are the obvious choice for automotive
Compared to graphics, ADAS asks different questions of the GPU, requiring something very different than traditional rendering. If we look at computer graphics, the use of compute shaders is now a standard for delivering advanced graphical effects. Essentially, the GPU runs small computer programs that define the colour and shade of the millions of individual pixels on screen. Rather than graphics, ADAS platforms can leverage this GPU compute capability to process and analyse sensor data in real-time.

And it’s not just sensors, but also conventional cameras that feed the GPU the data it’s traditionally been happy to work on. Image processing is a natural problem domain for the GPU. Indeed, almost any kind of computationally dense parallel computation is a good fit.


GPU virtualization for automotive
The level of performance future ADAS platforms will require necessitates increasingly larger GPUs, which will inevitably increase manufacturing costs. To mitigate this, platform vendors will look to increase the value of the GPU by using it to perform multiple workloads in the car.

This will only be possible if the GPU has rock-solid support for hardware accelerated virtualization.  Virtualization lets the GPU run multiple operating contexts, such as an app/OS container, without any of those contexts being aware of each other or in any way affect each other.

This is important. Imagine a situation where a problem with the dashboard software was able to affect the correct operation of the driver assistance system. This could be potentially catastrophic and must be avoided at all costs. The ability to have protected, virtualized, execution contexts supported by the GPU will ensure that this situation does not arise.

Virtualization works at its best when there is hardware support for entirely separate managed address spaces for each context to use and for restarting or flushing a context that’s misbehaving. This isolation is key to allowing cooperative use of the GPU, while keeping critical software, such as driver assistance systems, from being corrupted by any other process.


Enabling new revenue opportunities
From a car OEM’s point of view virtualization offers an additional benefit. It enables a safer environment to deliver various applications and services without any concerns about the electronics systems being taken down by a rogue piece of software.

It also means that rather than a traditional hardware box with fixed software for the infotainment and engine management systems the car becomes a flexible, configurable software platform that’s updateable over-the-air. It would enable OEMs to swap paid-for services in and out easily, without disruption to the operation of the car, thus offering them potential new revenue streams.

Why Imagination
Imagination’s PowerVR GPUs offer a set of desirable properties for ADAS functions. The core compute architecture inside today’s PowerVR GPUs was designed from the ground up to offer fast performance and low power consumption for reduced-precision computation, especially half-precision floating point (FP16).

Running at lower precision (where lower is usually classed as less than 32-bits) is one of the best ways to reduce power dissipation in an embedded GPU without significant loss of accuracy. We’ve designed the FP16 hardware as a separate data path from the full-precision FP32 hardware.

While shared data path designs are common since they’re simpler in many ways, having discrete hardware for each enables us to offer the best possible power consumption and efficiency because each data path accepts less compromise in the design to do what it needs to do.


As we have described virtualization enables all contexts running on the GPU to operate in isolation from each other to ensure that all ADAS functions are run safely, and as it is supported in hardware, it can do so at the performance level required. Hardware backed virtualization is a key strength of PowerVR and Imagination offers it on all its graphics core across the mid and high end. 

Imagination also delivers a toolset to support the development, optimisation and deployment of neural networks across GPU and AI accelerators. It is one unified tool-chain that enables developers to take multiple frameworks and multiple network types and bring them into a format that allows them to be deployed on either the GPU as a Compute Engine, on our PowerVR Series2NX and 3NX neural networks accelerators, or as a mixture of the two, where the flexibility of the GPU to implement a layer in a new variant of the network can be complemented by running the remaining layers on a highly optimised, high-performance dedicated CNN accelerator.

Thus, ADAS platform designers can use the GPU as a first-class component in the overall system architecture of the car, while allowing the system vendor to amortize their investment by also using the GPU to accelerate in-vehicle infotainment software, in a safe and secure manner.


The design of the PowerVR GPU microarchitecture, with its best-in-class power efficiency and memory bandwidth usage, provides a balanced GPU design that fits well with the car’s technology needs and lends itself perfectly for next-generation ADAS applications. It offers outstanding performance for the large and higher resolution displays with which the driver and passengers interact.

Looking to the future
The requirements of ADAS plays to the GPU’s inherent strengths, especially in the areas of image analysis and parallel signal processing. As these advanced driver assistance systems become more prevalent in the vehicles we use the importance of the GPU will only increase and systems that don’t utilise the GPU and function-specific accelerators to their advantage will be quickly left behind.

It is the GPU rather than the CPU that will be able to deliver the compute capabilities that will enable cars of the future to become more aware of their surroundings so that they can operate as smoothly and safely as we need them to.

About the author:
Bryce Johnstone is Director of Automotive Segment Marketing at Imagination Technologies – www.imgtec.com


Share:

Linked Articles
10s