The Right IP for Your Application
Across nearly every industry today, technology is advancing at an incredible rate, with just about every device becoming connected and intelligent.
Whether it’s in the cloud or in the edge, the systems-on-chips (SoCs) at the heart of tomorrow’s smart, connected devices will be designed with heterogeneous combinations of processors – including CPUs, GPUs, DSPs, AI processors and other accelerators. The idea is to make use of each processor’s specialized capabilities depending on the workload.
GPT provides the IP cores needed to design heterogeneous systems for connected devices at the IoT edge. For a broad number of applications, it’s important to run AI algorithms on edge devices (not in the cloud) to meet reliability, real-time response, efficiency, and security requirements. The only way to efficiently process the diverse compute workloads within the
power and size constraints of edge devices is through heterogeneous computing – running the various tasks on the most appropriate compute engine in the SoC. Heterogeneous computing architectures enabled by GPT’s Unity IP platform are able to provide more efficient, lower latency compute performance for performing inference on edge devices.
Because GPT’s Unity processor IP cores are highly scalable, low-power and performance-optimized, they are ideal for a wide range of applications across autonomous vehicles and intelligent transportation, smart cities, industrial automation, robotics, and machine vision.
Autonomous Vehicles and Intelligent Transportation
Technology for autonomous vehicles such as self-driving cars is advancing rapidly. As we move from L3 and L4 systems toward fully autonomous vehicles, what’s needed are systems that can process the huge influx of data from the multitude of cameras, LIDAR, and other sensors, in real-time with no delay. Parallel processing is critical for this task, and efficiency is a key consideration. The system must also decide how to respond to the data it’s been given, and for this, AI and vision processing processors are a must. GPT’s AI accelerator is an ideal fit for these applications.
Whether you’re designing a system for autonomous vehicles or another intelligent transportation application such as smart parking, intelligent transit, telematics, V2V communications, or others, you can select the Unity IP core or combination of cores that deliver the right solution.
Today we are only seeing the beginning of the smart city, with emerging solutions for resource management and ‘smart’ systems for buildings, parking, street lighting, surveillance cameras and more. Across the smart city, chips with intelligent processing will be critical.
In edge devices like smart cameras, performance requirements are increasing.
These cameras must support multidimensional and low-light image and video processing, powerful computer vision capabilities, 4K video encoding, communications capabilities, and a wide range of other features, all while providing the power efficiency and low cost needed for mass deployment.
GPT provides the Unity IP platform with IP cores such as the ‘Unity’ CPU cores, Variable Length Vector DSP (VLVm1) cores for signal processing applications, and ‘Song’ AI accelerators to deliver the powerful yet efficient solutions needed.
Industrial Automation and Robotics
To make production more efficient, automation is the way of the industrial future across factories, packaging facilities, utilities and power plants, and beyond.
Whether designing instrumentation, motion control, human machine interfaces (HMIs), distributed control systems, or other industrial automation systems, GPT’s Unity IP cores can provide the right processing resources with the cost efficiencies needed to further drive operational efficiencies.
In industrial and commercial robots and drones, heterogeneous processing systems can provide the real-time performance, small form factor, and low power consumption required. A heterogeneous system such as that provided by GPT’s Unity IP cores can enable drones and robots to recognize and track multiple objects and perform dynamic collision avoidance.
Over the past decades we’ve seen the emergence of image processors and sensors which can mimic the human ability of sight.
The technology at the heart of these systems continues to advance, improving in precision and sensitivity, and making it possible for cameras to quickly capture images (often with built-in depth perception), analyze raw sensor data in real time, and generate digital representations to aid in decision making.
Machine vision today is enabling a wide range of new use cases, which will only continue to increase. GPT’s Unity IP cores comprise the key processing blocks needed for machine vision applications.