core tide (ID: aichip001) | Xinyuan
core tide October 9th news, in the early morning of this year, at the Arm TechCon 2019 event held in San Jose, California, Arm announced the introduction of custom instruction functions in some CPU core , that is, customers can write their own customized instructions to accelerate their specific use cases , embedded and IoT applications.
Starting from 2020, all customers who use the Cortex-M33 core and later the Cortex-M CPU core series can use the custom command function for free.
▲Arm CEO Simon Segars discusses the new custom instruction set features on future Arm CPU cores.
Simon Segars said there are still many things to be arranged before the company returns to the open market. However, SoftBank CEO Masayoshi Son 's goal of returning to the securities market in 2023 remains unchanged.
At this conference, Arm also announced changes to its free open source IoT operating system Mbed OS partner governance model, allowing developers to have a more direct impact on the features of Mbed OS. In addition, Arm has established a self-driving car computing alliance and announced a collaboration with Unity to support various 3D content creation.
1. Introducing custom instructions into the CPU core
At this conference, Arm CEO Simon Segars announced that starting from Cortex-M33, Arm will integrate a special block in the CPU core, allowing customers to customize instructions on the Armv8-M instruction set architecture (ISA). Starting from 2020, all customers using the Cortex-M33 and thereafter, the custom command function will be available for free.
Arm Cortex-M core is a power-optimized kernel that can be used in conjunction with the Cortex-A performance core in applications such as smartphones and PCs. This core is also the main processing core of billions of low-power embedded/IoT applications, covering various terminal devices such as toasters, refrigerators, motor controls, sensor processing, etc., accounting for the vast majority of billions of Arm core shipments each year. Thomas Ensergueix, senior director of automotive and IoT business at
ARM, said Arm always has a range of acceleration options, starting with its memory-mapping architecture, connected via the bus and today’s neural processor units, which allows the CPU and accelerator to run in parallel, but at the same time the bus becomes a bottleneck, and customers can also choose a coprocessor connected directly to the CPU.
And today's newly released custom instruction feature allows Arm customers to create their own acceleration algorithms and then run them directly on the CPU, which means customers can integrate and quickly develop fully integrated custom CPU instructions faster without software fragmentation and not run in parallel like memory mapping solutions.
Because there are no interruptions in existing CPU capabilities, this setting provides lower costs and risks for integrating customer workload acceleration and still allows their customers to use existing standard tools they are familiar with or have used.
Currently, Arm has cooperated with many partners to support custom instructions, including IAR Systems, NXP, ChiNext Technology, ST STMicroelectronics, etc.
This is the second announcement released by Arm this year to make its products and technologies more flexible. Previously this summer, Arm announced a new licensing plan that provides customers with a lot of IP access at a lower price, without signing a license agreement before producing the chip.
With the help of Arm's flexible access services, customers can buy a single chip for only $75,000 a year, without paying $200,000 to purchase an unlimited number of chips.
These new measures help solve two limitations of the Arm architecture: one is the cost of developing chips, and the other is the flexibility of using unconventional hardware. The licensing changes and the ability to develop custom instructions make it easier for current and prospective customers to start developing chips with Arm IP while leveraging their advantages or Arm tools and a vast ecosystem.
2. Change the Mbed OS partner governance model
In addition to releasing custom command functions, Arm also announced that it will change its free open source embedded operating system Mbed OS.Mbed OS, as the basis of the Internet of Things (IoT), is helping everyday objects become smarter and more interconnected.
Over nearly 30 years, the Arm ecosystem composed of more than 1,000 partners has jointly developed a model that has delivered more than 150 billion chips, helping hundreds of partners to build billions of IoT devices based on Arm chips.
Arm says that the Arm ecosystem model is closely linked when solving the challenges of developing and deploying the Internet of Things.
In order to simplify IoT development, over the past decade, Arm has built a vast ecosystem around the free open source IoT operating system Mbed OS, which includes more than 425,000 third-party software developers, more than 150 Mbed-enabled motherboards and modules.
Currently, Arm is changing Mbed OS into a collaborative governance method, establishing a monthly product working group meeting, where Arm will prioritize and vote, and then add new features to Mbed OS. Any member of the Mbed Silicon Partner Program is allowed to participate.
Analog Devices, Cypress, Maxim Integrated, Nuvoton, NXP, Renesas Electronics, Realtek Semiconductor, Samsung, ChiNet Technology, u-blox and other Arm chip partners have actively participated in the working group.
▲Arm plans to become the adhesive for all technologies.
Previously, most commercial IoT operating systems were developed by a single vendor, without a clear model of how hardware partners contribute. Arm said the results of the product task force discussion will ultimately benefit the wider Arm IoT ecosystem to help silicon chip partners maintain strong business leadership and increase efforts to build new features, features and capabilities that are critical to scaling up to trillions of connected devices. At the same time, it also helps shape and determine the future direction of Mbed OS.
For example, Arm is performing new low-power battery optimization based on initial contributions from partners, which will extend the battery life of the device in the field.
As the scale of the Internet of Things market expands to hundreds of billions and trillions of connected devices, Arm said that only when partners can continue to cooperate can the ecosystem model succeed. This is an important step towards promoting continuous innovation and differentiation within the IoT ecosystem.
3. Establish the self-driving car computing alliance
At the conference, Dipti Vachani, senior vice president of Arm Automobile and Embedded Products, announced the establishment of the self-driving car computing alliance to solve various safety and computing problems in a collaborative way.
Alliance members include General Motors, NVIDIA, Denso, Toyota , Bosch , Arm, Continental and NXP.
Vachani said that many people are concerned about the safety of human driving and autonomous vehicles. In order to ensure that technologies from different providers work well, the alliance will cooperate on issues such as safety, computing, and software. She said no company has done this before.
4. Cooperation Unity supports 3D applications
At this conference, Arm announced that it will cooperate with Unity to ensure that 3D applications run smoothly on hardware using the Arm architecture. The announcement is part of a larger Total Compute collaboration approach in hardware and software design, where developers can easily access cores besides CPU cores. Arm Total Compute represents a new approach to IP design with a focus on use case-driven optimization system solutions.
Based on this solution, developers will write their software with the help of a software development kit, which can find out the best way to handle the software on CPU, GPU, or machine learning (ML) hardware, and learn how to optimize for optimal rendering and performance. Paul Williamson, vice president of Arm's client group, said it depends on the best way to handle the software within a given power range.
As part of Total Compute, Arm and Unity Technologies are expanding their strategic partnerships to further improve performance.
Williamson said Arm and Unity are expanding their collaboration to various 3D content, especially VR and AR, and it takes a lot of work to get it to its best on all platforms with computing constraints. The two sides have been working on engineering for several months and will continue to do so in the coming years. Arm's technology will be optimized for Unity's real-time 3D development platform and vice versa.
Unity Technologies vice president of platform Ralph Hauwert said he was pleased to work with Arm, which will ensure that the world's largest creator community - Unity developers - provides uniquely optimized tools for billions of Arm-powered devices.
▲Unity renders game and entertainment graphics.
Smythe believes that there are many transformations taking place from challenges in process expansion and data privacy to very decentralized ecosystems outside of CPUs. While 5G will create a world of opportunity, hardware manufacturers will still need to provide powerful functionality, performance and efficiency for the next generation of immersive experiences, and Total Compute will work hard to find a balance between 5G connectivity and the need to process screen graphics. With the advent of 5G, the acceleration of AI, xR and IoT is changing computing needs, and the performance required for digital immersion will have to exceed today’s level.
"This requires us to design intellectual property (IP) in a very different way, with the focus on optimizing performance, security, and developer access. There is no doubt that this will not be exactly the same, but just a shift in mindset." Smythe said, "By optimizing from within IP, software and tools and across IP, software and tools to provide a secure foundation and provide the performance needed to address the complex computing challenges of future workloads, we are shifting from a single product development focus to use cases and experience-driven system solutions."
V. Continuous improvement of machine learning performance and security capabilities
Since Arm introduced the Cortex-A73, Arm has gradually improved the performance of machine learning (ML), and today, Arm hopes to significantly expand the use of ML in CPUs. "To implement this new digital world, we need to push computing to a higher level, which is why we added Matrix Multiply (MatMul) to our next-generation Cortex CPU Matterhorn, thus making ML performance more efficiently doubled generations than before."
According to him, in addition to CPU, Arm needs to focus on the Total Compute method and apply it to every computing element in the system and its infrastructure. Whether it is an Arm CPU, GPU, NPU, interconnect module or system IP, it must be optimized as an integrated solution. This depends on the software and tools that achieve this goal, including Arm NN, Arm computing libraries, open source communities and open standards, etc., which are all built on security.
▲Ian Smythe announces the efforts made by Total Compute.
As people are increasingly relying on mobile phones as the center of all personal information, privacy and security matters, Arm has now begun launching innovative security features such as Storage Tag Extension (MTE) as part of Total Compute to meet the needs of customers.
Arm hopes that Total Compute will become the preferred platform for all developers.
Recently, Google announced plans to collaborate with Arm to design MTE for Android devices. These capabilities, combined with Platform Security Architecture (PSA), will help standardize and defragment the security of the entire ecosystem.
Smythe said any Total Compute solution involves a lot of elements, whether it’s for VR headsets or wearables, smartphones or DTVs, with the goal of providing the foundation for future computing platforms and adopting a comprehensive computing approach that simplifies security, improves performance and efficiency, and provides developers with more performance access opportunities throughout the Arm ecosystem, ultimately enabling a truly digital immersive experience.
Conclusion: In line with the general trend of developing customized chips
In the past year, the rise of open source instruction set RISC-V and the announcement of open source of old MIPS have all launched an impact on Arm. From Arm's series of measures since this year, we can see that Arm is rapidly adapting to market changes and eliminating competitive pressure.The newly released custom instruction function of
solves a major limitation of using third-party IP cores and caters to the ever-growing trend in the electronics industry, namely developing customized chips to optimize performance and reduce costs. For Arm, this is an important step to embrace customization of instruction sets while still maintaining the security, predictability and reliability of its kernel. Developing customized chips with
still requires huge costs and risks, but as this threshold gradually decreases, more and more companies such as consumer electronic devices and automotive OEMs are developing or considering developing customized chips internally or through partners.
Original text from: VentureBeat, Forbes, Techcrunch
