Arm offers bespoke silicon and closer ties to drive innovation and scale in IoT
In a trio of announcements at Arm TechCon 2019 in San Jose yesterday (October 8), chip design company Arm walked the line on increasing flexibility and reducing fragmentation in the IoT space, as the rise of 5G and AI place ever-higher demand on compute power, hardware-software integration, and easy scalability.
Most significantly, Arm is enabling silicon vendors to tailor the hardware in Arm-based chip designs for the first time for specific IoT applications. The idea is to allow for customisation of system-on-chip (SoC) design to drive scale and performance of IoT deployments without causing software fragmentation.
The UK-headquartered company said greater flexibility in IoT design, enabled by closer integration of hardware and software at silicon level, is required for the market to reach one trillion IoT devices, as it predicts for 2035. Arm Custom Instructions, its new offer for SoC designers will allow bespoke hardware-software optimisation with the Armv8-M architecture.
The Armv8-M architecture covers design of its real-time ARMv8-M embedded processors, for low cost embedded IoT systems, going all the way “from the smallest IoT device to complex SoCs”. Arm Custom Instructions will available in Arm’s Cortex-M33 CPUs, used in microcontrollers and shipped in many billions, starting in the first half of 2020.
The move to allow partners to customise hardware design, as prescribed in Arm blueprints, in order to optimise performance of the software applications running on top marks a significant move for the company, which has made its original design ‘architecture’, governing the inter-working of Arm-based silicon, sacrosanct and untouchable.
But it said the new approach will afford chip designers flexibility to push performance and efficiency by adding their own application-specific features into Cortex-M33 CPUs. Partners will be able to interleave their own instructions with standard Arm instructions, it said, enabling workload acceleration and unlocking differentiation for strategic use cases driven by IoT, AI and 5G.
Dipti Vachani, senior vice president and general manager for Arm’s automotive and IoT bBusiness, said: “A world of a trillion secure intelligent devices will be built on a diversity of complex use cases requiring increased synergy between hardware and software design. We have engineered Arm Custom Instructions to fuel closer hardware and software co-design efforts toward achieving application-specific acceleration while unlocking greater device differentiation.”
The company explained the new facility is enabled by modifications to the CPU that reserve encoding space for designers to add custom datapath extensions while maintaining the integrity of the existing software ecosystem. “This feature, together with the existing co-processor interface, enable Cortex-M33 CPUs to be extended with various types of accelerators optimised for edge compute use cases including machine learning and artificial intelligence,” it said.
At TechCon 2019, Arm continued with its flip-side tale of fragmentation and integration in the IoT space, as the market is buffeted and swelled by the rise of 5G networks, artificial intelligence (AI), and other compute-hungry digital technologies.
Notably, it announced a deal with video game company Unity Technologies, maker of Fortnite, to optimise performance on Arm-based SoCs, CPUs, and GPUs. Fortnite and other Unity titles account for 70 per cent of VR content development today, it said. The deal will leverage its newly announced ‘total compute’ strategy, which seeks to drive deeper integration of AI in its processor design.
Ian Smythe, vice president of marketing for Arm’s client line of business, said: “While 5G will create a world of opportunity, we still need to deliver the power, performance, and efficiency for the next generation of immersive experiences.
“Along with 5G, the acceleration of AI, xR, and IoT are changing compute requirements. The performance we need for digital immersion is going to have to push beyond what we have today, towards the world of Total Compute. This requires a very different approach to how we design our IP, with a deep focus on optimizing performance, security, and developer access.
“Make no mistake, it won’t be more of the same, but a paradigm shift in thinking. We are moving from a single product evolution focus to a use-case and experience-driven system solution by optimising within and across IP, software, and tools to provide a secure foundation and deliver the performance required for the complex compute challenges of tomorrow’s workloads.”
Arm is broadening its CPU coverage for machine learning (ML), it said, “to push compute to a higher level”. It is adding Matrix Multiply (MatMul) to its new Cortex “Matterhorn” CPU, doubling ML performance over previous generations. It is also seeking deeper integration of AI in sundry processors, interconnect, and system IP, via its Arm NN inference engine and Arm Compute Library for computer vision and machine learning.
Its total ‘compute strategy’ puts renewed focus on security, based on three layers: by rolling out of features like Memory Tagging Extension (MTE), which Google has said will be designed into Android devices; by pushing standardisation in software and driver interfaces in IoT platforms; and by tackling the general issue of fragmentation in the IoT space.
Smythe commented: “Our industry is reaching a pivotal crossroads as we try to solve a significant integration challenge: individual IP and fragmented solutions are hard to optimise. If we are to take advantage of the enormous future technology opportunities, we need to put use cases, consumer experiences, and the needs of the ecosystem at the heart of our IP design.”
Meanwhile, Arm has also introduced a new Mbed partner governance model, again geared towards driving innovation and scale in the IoT space. The new model is intended to offer silicon vendors a closer say in the development of the open-source Mbed operating system (OS). Arm said it will convene monthly working group s with silicon partners to “prioritise and vote on” new capabilities to be added to the Mbed OS.
“Our silicon partners can directly influence and enhance our efforts in building out new capabilities, features and functionality, which are critical in scaling to a trillion connected devices,” said Chris Porthouse, vice president of product and operations within Arm’s IoT services group.
He added: “Our new governance approach is unique… Most commercial IoT operating systems are often developed by a single vendor without a clear model on how hardware partners can contribute. The outcomes of [these]p discussions will ultimately benefit our broader IoT ecosystem.”
The likes of Analog Devices, Cypress, Maxim Integrated, Nuvoton, NXP, Renesas, Realtek, Samsung, Silicon Labs, and u-blox are already involved, and have contributed new work on low-power battery optimisation.
The TechCon 2019 announcements were appended by a series of partner quotes. The best of these are below.
Speaking of the Custom Instructions initiative, Geoff Lees, senior vice president and for microcontrollers at NXP, said: “Arm’s new Custom Instructions capabilities allow silicon suppliers to offer customers a new degree of application-specific instruction optimisations to improve performance, power dissipation, and static code size for new and emerging embedded applications.”
Alessandro Piovaccari, chief technology officer at Silicon Labs, added: “The world of one trillion securely connected devices poses several challenges to silicon designers due to the expanding demand for extremely energy-efficient devices. The introduction of Arm Custom Instructions will enable systems uniquely optimised for specific tasks within a wide range of connected IoT devices.”
Ricardo De Sa Earp, general manager for microcontrollers at of STMicroelectronics, said: “The relentless demand for greater performance, power efficiency, and security require an evolution of the design approach where hardware is designed alongside software from the beginning.”
Speaking of the Mbed OS announcement, Jason Lin, vice president for microcontroller applications at Nuvoton, said: “The move to open governance will continue to spur innovation on the key features and capabilities that our customers are requesting.”
Daryl Khoo, vice president of marketing for the IoT platform business at Renesas, added: “The new governance model gives greater opportunity to determine necessary features in order to future proof our offering.”