Arm’s role in IoT is changing

For embedded engineers, they must have different degrees of understanding of Arm kernel programming, whether it is STM32-based MCU, Raspberry Pi-based Linux development, or Rockchip-based Android development board, I believe everyone can have fun .

For embedded engineers, they must have different degrees of understanding of Arm kernel programming, whether it is STM32-based MCU, Raspberry Pi-based Linux development, or Rockchip-based Android development board, I believe everyone can have fun .

In order to accelerate the development of the Internet of Things, Arm launched a strategy called Total Solutions for IoT half a year ago. Networked development costs, product development cycles, and fragmented applications are supported, and through high-level synthesis, the entire development process from chip to embedded is simplified and accelerated.

A few days ago, Arm announced a series of major innovations in the Internet of Things, including the latest Corstone subsystem, the latest Arm Cortex-M85 that takes into account security, scalar performance and ML performance, and a series of extended support for Arm virtual hardware, and officially supports China , available through the Amazon Cloud Technology Marketplace (China).

It can be said that Arm has gone far beyond the scope of IP suppliers in the traditional sense through comprehensive IoT solutions.

Cortex-M85 comparable to Cortex-A

As the latest Cortex-M85 of the Armv8.1-M architecture, a picture can illustrate its unique scalar and ML performance.

As shown, Arm’s Helium technology enables integrated pointer authentication and branch target recognition without the need for a DSP. The scalar performance of the M85 is 30% higher than that of the M7, the ML performance is 20% higher than that of the M55, and it also supports TrustZone security.

The detailed indicators are shown in the figure above. Some of the more distinctive ones include the performance of 3.13 DMIPS/MHz and 6.28 CoreMark/MHz, far exceeding the Cortex-M7, and will support dual-core lockstep in the future. features. In addition, for the first time, PACBTI (Integrated Pointer Authentication and Branch Target Identification) is supported, simplifying the process for developers to achieve PSA-certified Level 2 security. PACBTI provides additional protection against return-oriented and jump-oriented software attacks by authenticating function calls and return addresses.

The Cortex-M85 uses a more advanced memory system architecture to ensure higher data and code throughput. A low-latency memory system with Tightly Coupled Memory (TCM) ensures deterministic operation. Four 32-bit wide data TCM interfaces and one 64-bit wide instruction TCM interface are available, all with integrated ECC. An additional 32-bit AHB access interface port allows simultaneous access to the TCM by an external DMA controller.

The Cortex-M85’s memory system L1 cache, also with ECC, connects to external memory via the AMBA 5 AXI master interface to optimize performance during slower, indeterminate memory accesses.

“It brings the performance of Cortex-M closer to the performance we have when using Cortex-A, but with the convenience, determinism and low power consumption of MCU.” Mohamed Awad, vice president of Arm’s IoT and Embedded Business Unit, commented. road.

Two latest Corstone subsystems

Arm Corstone, a high-level package of CPUs, NPUs, system IP, and other key technologies into a pre-validated and integrated hardware subsystem, Corstone is built for specific use cases and targeted directly to chip design partners and OEMs. By providing pre-verified, pre-integrated IP suites, chip manufacturers do not need to develop designs related to the Arm processing part, so they can complete chip development more quickly.

This update includes Corstone-1000, a comprehensive solution for Arm cloud-native edge devices, and Corstone-310, a comprehensive solution for Arm speech recognition.

Among them, Corstone-1000 integrates Cortex-A32 and Cortex-M, so it can not only use Linux and other operating systems, but also realize application-level workload development on devices such as smart wearable devices, gateways, and high-end smart cameras. The Corstone-1000 is Arm SystemReady™-IR compliant and has a PSA Certified hardware security enclosure for higher security levels, so OEMs can immediately benefit from Project Cassini.

Corstone-1000 easily implements Arm Cortex-A and Cortex-M heterogeneous systems, meeting the high performance and high energy efficiency of a single system. Cortex-A processors are scalable from Cortex-A32 to Cortex-A53, offering a variety of application options such as smart wearables, gateways and high-end smart cameras. The Corstone-1000 comes with a range of well-proven configuration options.

The Arm SystemReady IR certification of the Corstone-1000 facilitates software reuse to help further reduce time to market. Arm SystemReady IR is a hardware and firmware compliance certification program that enables common off-the-shelf operating systems to work out of the box, making one-time software development efforts scalable.

The Corstone-1000 is equipped with two prototyping platforms to enable early software prototypes. A Fixed Virtual Platform (FVP) and an FPGA-based system, both supported by an open source reference software stack.

Security has been fully integrated since the definition of Corstone-1000. The Secure Enclave (SE) is an important part of it. It uses the Cortex-M0+ processor, which is included in the Corstone-1000. SE has independent system resources and independent operating system, which can be used as a secure environment to store confidential and process sensitive information. The Cortex-M0+ in SE is the Root of Trust (RoT) for the SoC. The open source software stack consists of two bootloading stages on a Cortex-M0+ based on mcuboot and TF-M as runtime software. The main processor Cortex-A is controlled by SE. The Corstone-1000 is PSA Level 2 certified thanks to its SE.

The Corstone-310 integrates the Cortex-M85 and the optional Arm Ethos-U55, making it Arm’s current highest-performance MCU-level processor with AI capabilities. The solution targets use cases such as smart speakers, thermostats, drones, factory robots, and more.

Effectively integrating processor IP into a system can be difficult. A common challenge is to build a power and clock control system to meet the application’s stringent power consumption. The subsystems included in the Corstone-310 have pre-integrated power and clock domains for precise control of different aspects of the design.

Security is another key challenge for SoC designers. The Cortex-M85 is a big improvement in this regard, supporting Arm TrustZone and PACBTI. However, extending this level of security to other parts of the system can present additional challenges. Building a complete security infrastructure can be a time-consuming task, and mistakes can lead to serious security breaches. The Corstone-310 subsystem has a pre-built trust zone infrastructure – enforcing secure and non-secure code isolation throughout the system, including partitioning memory and designating peripherals as secure or non-secure.

Additionally, Corstone-310 meets many of the requirements of TBSA-M, a security best practice specification when designing systems around the Armv8-M architecture. TBSA-M is part of Arm PSA, so Corstone-310 provides an excellent starting point for achieving PSA certification. Secure software design is also simplified thanks to the Trusted Firmware-M (TF-M) port, which enables a secure processing environment for Armv8-1 M and Armv8, paving the way for PSA certification.

Awad said: “Alif semiconductor has entered the mass production stage with Corstone. Since Arm Corstone was released three years ago, our customers have now adopted this technology in more than 200 design projects.”

Virtual Hardware Expands Further

Arm virtual hardware is a transformative product that allows developers to directly open the IDE, connect to the desired hardware virtualization platform, and start software development immediately without waiting for the chip and development board to be fully completed.

It enables the Arm ecosystem to easily cloud-based development and continuous integration/continuous development (CI/CD) without the need for large custom hardware clusters. Arm virtual hardware is already used by hundreds of developers. Based on their feedback, Arm has also added several new virtual devices to expand the appeal of Arm virtual hardware, including Arm virtual hardware designed for the new Corstone, and seven Cortex-M0 to Cortex-M33 Cortex- Arm virtual hardware for M processors. In addition, Arm is further expanding its technology library with third-party hardware from partners such as NXP Semiconductors, STMicroelectronics and Raspberry Pi, including Raspberry Pi 4, NXP i.MX 8M and STM32U5.

Extending Arm virtual hardware to ecosystem devices and the vast majority of Cortex-M products, ISVs and cloud service providers can now leverage the billions of deployed Arm architecture-based IoT and embedded devices, and There is no need to purchase various development boards, reducing maintenance and development costs.

Currently virtual hardware can support three Corstone platforms, including Corstone-300, Corstone-310 and Corstone-1000. Arm virtual hardware Corstone is available as an Amazon System Image (AMI) on AWS Marketplace, and Chinese developers can also access the Chinese version of AWS Marketplace. The virtual hardware currently supports seven Cortex-M processors, including Cortex-M0, Cortex-M0+, Cortex-M3, Cortex-M4, Cortex-M7, Cortex-M23 and Cortex-M33. Arm virtual hardware has been integrated into Keil MDK, making it a natural part of the development process in traditional embedded development workflows, eliminating the need to select a development board in the IDE, just select virtual hardware.

Developers can now seamlessly move software testing workloads to the cloud and scale to hundreds or thousands of virtual boards in minutes, dramatically speeding up the software development process. “First, developers like being able to write software for hardware before it’s available, but they also want to be able to access and extend their software across the billions of IoT devices in the field. Second, developers like AMIs that allow them to Integrate Arm virtual hardware into their development process as they see fit, but they don’t want to change their process. They tend to integrate Arm virtual hardware directly into the existing tools and services they use every day. Finally, in China, the ecosystem Local support is required. Both to ensure that the Arm virtual hardware target is easily accessible, but also to ensure that local AI partners can easily access and utilize the technology.” Awad told EEWORLD reporters.

Over the past few months, many leading Chinese AI companies have heavily adopted Arm virtual hardware, such as Baidu, Shenzhen VoxAI (VoxAI), Shanghai Max (mSlink), and Sound+.

Open IoT SDK Framework

To scale, the Cortex-M software ecosystem needs a consistent set of standards for portability and reuse of software across multiple devices. Project Centauri was born to do just that, allowing developers to focus on the innovation and differentiation that really matters. Open-CMSIS-Pack has been supported by 9,500 microcontrollers and 450 development boards since its launch last year, allowing software vendors to easily expand their products on these devices. Open-CMSIS-Pack is a Common format used in the developer environment. The most popular real-time operating systems, such as FreeRTOS and TencentOS, are added to the list of available software in CMSIS-Pack format. At the same time, it has also started to integrate CMSIS-Pack functions into the most popular IDEs, such as VSCode and so on.

Arm is also releasing the first version of the Open IoT SDK Framework, which includes the new Open-CMSIS-CDI software standard, a community-driven project hosted in Linaro that defines the Cortex-M ecosystem a generic device interface. Eight major industry players from chip partners, cloud service providers, ODMs and OEMs have joined so far.

“Project Centauri is just getting started, but the momentum is incredible, and I’m looking forward to seeing it reach its full potential,” Awad said.

Long-term goals of software and hardware co-design at the system level

As shown in the figure, in order to accelerate the development of IoT from chips to software applications to product-level development, Arm has launched a comprehensive solution, from processor IP to Corstone clusters, to virtual hardware and a broad software ecosystem, and thus gained Recognized by multiple partners. Awad also affirmed the many breakthroughs achieved in the six months since the comprehensive solution was launched.

He said: “In the next ten years, we estimate that IoT hardware will grow at about 10% annual growth rate, and software plus services will grow at about 20% annual growth rate. That weight will ultimately outweigh the hardware, which is why Arm spends so much time making sure the software works well and making development on Arm architecture easier.”

Customer Reviews

Arm’s success in the IoT field is inseparable from its partners and ecosystems. The following are the partners’ recognition of Arm’s comprehensive IoT solutions.

Richard Barry, Senior Principal Engineer, IoT at Amazon Cloud Technologies and Founder of FreeRTOS: “FreeRTOS is the most popular RTOS for embedded development, supporting hundreds of different Arm devices. This broad device support capability requires efficient continuous testing posed a challenge today with clusters of boards that are expensive to operate. Arm virtual hardware running on the AWS cloud provides an easier, faster and more cost-effective way to scale our CI testing.”

Zhao Qiao, head of Baidu’s PaddlePaddle product team: “In the era of the Internet of Things, the efficient integration of deep learning open source open platforms and chip design platforms will greatly improve the efficiency of industrial developers in developing smart devices. And the inference capability is deeply integrated with the Arm virtual hardware platform. Developers can quickly and efficiently deploy the most popular PP series industrial-grade featured models on Github on the Cortex-M series hardware on the cloud without completing the chip. Prototype verification of the system.”

Tom Trahan, VP of Business Development at CircleCI: “The new generation of IoT applications will require the adoption of modern development practices such as continuous integration and continuous delivery (CI/CD), etc. Through Arm virtual hardware, CircleCI’s on-demand runners can dynamically provision EC2 Example, enabling IoT developers to scale their development and testing without owning physical hardware. Combined with Arm virtual hardware, CircleCI radically reduces the time from idea to implementation for Arm architecture-based IoT applications.”

Zach Shelby, co-founder and CEO of Edge Impulse: “Finding the right balance between DSP configuration and model architecture and memory and latency constraints is a major challenge for edge ML developers. By accessing more styles of Arm With virtual hardware, we use the cloud to easily estimate the performance of more IoT devices. This will ultimately shorten developer turnarounds and give them the confidence to deploy optimized models to actual edge devices.”

Clay Nelson, VP of Sales at GitHub: “Millions of developers and businesses worldwide use GitHub for cloud-based software development for their wide variety of applications. The combination of Arm virtual hardware and GitHub Actions enables IoT and embedded Developers can do CI/CD without the need for physical hardware.”

Nima Badiey, VP of Global Alliances at GitLab: “We believe that the increasing pace of IoT software development is on par with the Internet and mobile applications. IoT developers need an end-to-end DevOps solution to rapidly develop, test and deploy new Connected Services. Arm virtual hardware enables developers to test software for Arm-based IoT devices directly in the cloud and build on GitLab’s proven enterprise DevOps platform, effectively leveraging the same development used by millions of developers , continuous integration and delivery tools.”

Nico Mäding, Manager SoC/ASIC Design at Hilscher: “The Arm Corstone-1000 provides the foundation for a highly secure SoC platform, allowing us to focus our design resources on developing flexible programmable IP for the future industrial automation market Demand for Gigabit Ethernet networks powered by TSN. Arm and Hilscher are working together to enable highly secure embedded software designs for next-generation industrial IoT communication solutions.”

Guo Shengzhe, Associate Director of Qijing Optoelectronics: “Edge Perception, the customized image processing chip solution of Qijing Optoelectronics, is based on the Arm Corstone-300, which is widely used and recognized by the industry, and runs seamlessly on the Arm Virtual Hardware Target (AVH). , making our SoC more compatible with the hardware and software support architecture of the Arm ecosystem. And AVH is about to expand to third-party device platforms, which will make the development process smoother. We also plan to import Project Centauri to make our chip solutions compliant with the basic The standards and security requirements. These will greatly shorten the development process and time for developers, focus on specific usage scenarios, and improve customer experience.”

Chen Ran, CEO of GitLab: “GitLab is committed to providing users with an open and integrated DevOps platform covering the software development life cycle. Combined with Arm virtual hardware, the GitLab platform further enhances the value of DevOps and empowers development It enables users to deliver software faster, while helping companies accelerate the introduction of innovative IoT products.”

Wang Wei, Technical Director of Shanghai Max: “Artificial intelligence will become a key force in the development of IoT applications in the future, and more and more professional chips for the field of machine learning will continue to be born. Arm virtual hardware enables us to target the new Cortex-M55 and Ethos-U55 conducts software library development and testing, and supports these new processors before the chip is completed. Especially in the design of artificial intelligence models, enabling developers to get rid of the constraints of hardware and evaluate and debug models in the cloud in advance performance running on a chip, resulting in significant savings in product design cycles.”

Cristiano Castello, senior director of product innovation for edge processing microcontrollers at NXP, said: “The ability to scale microcontroller applications to deliver higher performance computing, security and enhanced user experience is critical to meeting the growing demands of our customers. NXP is pleased with the addition of the Cortex-M85 to the Arm Cortex-M processor family, which will enable new embedded product applications. NXP is also pleased with our popular i.MX 8M Plus Core Complex-based development board Now available in the cloud as Arm virtual hardware instances. Our customers can now automate their application development process and maximize the scalability of those applications before they are deployed on physical hardware.”

Leo Leung, VP of Product Management at Oracle: “Oracle Cloud Infrastructure Ampere A1 Compute instances enable software developers to run continuous integration workloads efficiently and at scale. With the new Arm virtual hardware from Oracle Cloud Marketplace, IoT development will be reduced complexity without compromising performance and security.”

Sang Won Lee, co-founder and CEO of Qeexo: “Qeexo has developed machine learning solutions that generate actionable insights from sensor data. By integrating Arm virtual hardware into our platform, customers now do not need physical hardware The ability to run and measure the performance of multiple Qeexo machine learning models simplifies machine learning development for Arm-based IoT devices.”

Eben Upton, CEO of Raspberry Pi: “The Raspberry Pi puts computing power in the hands of people around the world. Providing an easily accessible virtual version of the Raspberry Pi through Arm virtual hardware will allow more developers to test We technology, solve problems and express ourselves through creative projects.”

Sailesh Chittipeddi, Executive Vice President and General Manager of the IoT and Infrastructure Business Unit, Renesas Electronics: “Thanks to the advantages of the Arm Cortex-M platform, the Renesas Electronics RA microcontroller series has achieved extraordinary market performance, setting a new record for Renesas Electronics. A new record for the fastest adoption of the new MCU family in electronics. As we continue to expand our high-performance and advanced IoT security MCU business, Renesas welcomes the timely introduction of the Cortex-M85, a processor that provides an industry breakthrough In addition, Arm virtual hardware will complement our rich software development ecosystem to expand the reach of the platform and provide a great user experience for developers around the world.”

Qiu Fenghai, CEO of Sound Plus Technology: “The wearable device and IoT market is developing rapidly, and it is necessary to apply machine learning methods to the terminal device to complete the calculation of audio. Arm’s strong ecosystem and flexible IoT virtual hardware are conducive to the development of online algorithms. Rapid verification and iteration, compressing the product development cycle. Based on the software-hardware solutions provided by Arm IoT solutions, we provide customers with advanced functions such as voice enhancement, active enhancement, and acoustic scene recognition based on machine learning method updates, and provide customers with advanced functions such as voice enhancement, active enhancement and sound scene recognition. of consumers provide an enhanced audio experience.”

Ricardo De Sa Earp, Executive Vice President, General Microcontrollers Sub-Product Group, Microcontroller and Digital IC Products Group, STMicroelectronics: “Developers who want to push the boundaries of the IoT and embedded markets need access to the right tools and technologies. Arm virtual hardware disrupts the market for developers, enabling easy transfer of software from virtual models to hardware. Adding the high performance and TrustZone capabilities of the Cortex-M85 core to STM32 MCUs is an important opportunity for developers to push the boundaries of new interconnect and security applications, And will open up a whole new range of AI use cases.”

Deqiang Chen, Software Engineer, TensorFlow Lite For Microcontrollers: “Given the growing demand for machine learning at the edge, we are partnering with the latest Arm technology through TensorFlow Lite For Microcontrollers. Expanding Arm virtual hardware to include new virtual devices and new Cortex-M models expand This extends the reach of our cloud-based continuous integration system and provides the opportunity to develop machine learning applications using TensorFlow Lite For Microcontrollers without access to physical hardware. This is an important enabler to accelerate the development of innovative machine learning applications at the edge.”

Liu Aifeng, CEO of Shenzhen Weiai Intelligence Co., Ltd.: “The voice AIoT market has demands for the implementation of diverse algorithms in various sub-scenarios. In the past, our algorithm team needed to compile and adjust each algorithm separately for each chip platform. Excellent, this workload is extremely huge. Arm virtual hardware technology can help us build a normalized and standardized platform. On this platform, we can integrate various algorithms such as voice wake-up and command, voice noise reduction, and voice event recognition. It can be adjusted to the optimal state and can be adapted to various hardware solutions based on the Arm architecture, which greatly shortens the time for the algorithm from demand to commercial use, and also reduces related costs.”

The Links:   CM150DC1-24NFM 1MBI400S-120