At the first Dark Iron RISC-V Ecological Conference held by Alibaba Pingtou on March 2, Ni Guangnan, an academician of the CAE Member, said that today RISC-V is the most popular architecture in China's CPU field and a new engine to promote the development of new generation information technology.

David Patterson, the father of RISC-V and winner of the Turing Award, also praised the thriving development of the RISC-V ecosystem. He said, "RISC-V is a global phenomenon that has been studied by developers from over 60 countries. In Asia, many well-known enterprises, academic institutions, and industry associations are actively involved. From embedded systems to various types of computers, and finally to mainframes, I believe that in 3-5 years, RISC-V will be everywhere

According to the data, RISC-V architecture, which started in 2010, is an open source, free instruction set architecture based on the principle of reduced instruction set (RISC), and is also regarded as the third largest CPU architecture that can impact x86 and Arm. It is fully open, low-cost R&D, compact and flexible. The industry points out that 10 billion processors using RISC-V architecture have been shipped in 2022, and it is expected to exceed 80 billion by 2025.

Based on the above advantages and promising development prospects, the RISC-V architecture has been widely favored by technology giants such as Alibaba in recent years. Pingtouge is a wholly-owned semiconductor chip business entity of Alibaba Group, announced in September 2018. It is one of the earliest companies to deploy RISC-V technology and continues to promote the development of RISC-V.

What are the latest developments in technology, applications, and ecology in the RISC-V field since its establishment 5 years ago? How does the industry view RISC-V and the development opportunities and challenges behind it? At the conference that day, Pingtou Brother and his ecological partners were interviewed and provided answers.
Ping Tou Ge: The Prosperous Development of RISC-V Industry Needs Global Innovation Collaboration

Meng Jianyi, Vice President of Ping Tou Ge, introduced that in the research and development of RISC-V processors and chips, Ping Tou Ge has launched three major series and eight RISC-V processors, covering scenarios such as high performance, high energy efficiency, and real-time reliability. It is the most widely used processor in RISC-V mass production chips.

In terms of operating systems, Pingtouge has basically completed the full adaptation of international and domestic mainstream operating systems with RISC-V, including Android, Debian, Fedora, Gentoo, Ubuntu, Dragon Dragon Dragon, Tongxin, openKylin, Skyworth Kukai System, RTT and other operating systems. This means that RISC-V has the full ability to undertake software applications through different operating systems such as cloud, desktop, and terminal. Upper level application developers can smoothly achieve architecture compatibility with 0 code.

At the software tool and application level, Pingtouge has for the first time fully launched a basic software three piece set - compiler TAC, compilation environment CDK, and deployment tool set HHB, greatly improving the efficiency of RISC-V development environment and tool stack. At the same time, based on the adapted openKylin operating system, RISC-V has for the first time become a staple of large-scale commercial IM (instant messaging) tools.

At the same time, Alibaba Pingtou announced its ecological plan. For 150000 developers in the chip open community, Pingtou Ge has launched its first online learning platform, the RISC-V Academy Program, to encourage participation in global RVFA certification. For industry partners, Pingtou Brother has launched the "Xuantie Preferred Chip" program to promote the implementation of ecological chips; At the same time, we have set up the "Xuantie Preferred Partners" and awarded the first batch to 8 companies, including Quanzhi Technology, Apt, Compute Energy, Boliu Intelligence, Qingwei Intelligence, Yunzhisheng, Jiangxin Chuang, and AFP IoT, in recognition of their outstanding contributions to the RISC-V ecosystem.

The vigorous development of RISC-V requires global innovation and collaboration. From the chip ecosystem, to the basic software ecosystem, and then to the application and terminal ecosystem, Pingtou Ge is connecting various ecosystems, allowing developers and partners around the world to better use and develop RISC-V technology, "Meng Jianyi said at the conference.
Dragon lizard: RISC-V based solutions will have a significant impact on the operating system ecosystem

Ma Tao, the chairman of the Dragon Lizard Community and researcher at Alibaba Cloud, introduced the application and future actions of the Dragon Lizard operating system based on the Flat Head Brother RISC-V chip.

Last year, Alibaba Cloud and Pingtou Ge, as well as the PLCT laboratory of the Software Institute of the Chinese Academy of Sciences, established RISC-V ARCH SIG, mainly engaged in work based on RISC-V distributions. And at the RISC-V China Summit that year, they announced their first cooperation achievement - a full stack platform of software and hardware, ranging from chip platforms to Dragon Lizard operating systems to application software. This platform can cooperate with the Wujian 600 platform released by Pingtouge and the Zhaoying 1520 platform, and can run some office applications very smoothly.

From user feedback, it can be seen that RISC-V is widely recognized for its capabilities in various aspects, and there are also many industry colleagues working together to build RISC-V capabilities at Dragon Liza RISC-V ARCH SIG.
He predicts that the RISC-V ecosystem or various solutions based on RISC-V will have a significant impact on Alibaba Cloud and even the entire server operating system ecosystem in the future. Therefore, Dragon Lizard is also continuing to promote work in this area.

Dragon Lizard has also made many attempts to promote RISC-V related technologies, and the review work of the recently published book "RISC-V Architecture Programming and Practice" has involved the Dragon Lizard community. I hope to promote the development of RISC-V and RISC-V based operating systems through these efforts.
Ma Tao stated that an important task of operating systems is to adapt to various chips. The booming development of RISC-V is a huge opportunity for the operating system community, manufacturers, or practitioners working in the operating system industry.

In the future, if RISC-V is used in various scenarios of "cloud edge end", firstly, it can make the operating system more universal and universal, such as using a set of software to solve problems in various scenarios; The second is that operating system developers can delve deeper into the development of more valuable key components, which is very beneficial for many open source communities and operating system development.
Calculation: Actively exploring the combination of RISC-V and AI

Wang Lei, Senior Vice President of Quanneng, discussed the relationship between RISC-V and AI at the meeting, stating that Quanneng has also invested in server chip development and AI chip development. ChatGPT and a series of other applications from last year have made 2022 a very important year for AI development, establishing very clear application requirements, especially for AI hardware.

In the field of accelerators, Quanneng is also actively exploring and contemplating the combination of RISC-V and AI. Wang Lei stated that this combination is a very natural choice. At present, AI, whether for training or reasoning, may be more based on GPU, and GPU itself is a relatively specialized accelerator, and there is still a collaboration problem between GPU and its main CPU.

As the model size increases, CPUs that are closer to data flow will form a deeply coupled architecture with acceleration engines, and further improvement in coupling efficiency is needed. RISC-V points out a very good direction, as it has strong scalability and can effectively solve latency and bandwidth issues between CPU and acceleration engines, forming a more complete and consistent solution and structure.

Wang Lei stated that Quanneng is also exploring this direction, and he has observed that Google's TPU is also developing a programming paradigm that combines acceleration engines and RISC-V to support AI workloads with higher efficiency.
Yun Zhisheng: ChatGPT is a significant benefit for the server side

Li Xiaohan, co-founder of Yunzhisheng, discussed topics such as cooperation with Pingtou Brother and his views on ChatGPT and AI at the meeting.

He stated that there are two reasons for collaborating with Pingtou Brother. Firstly, more and more partners are using RISC-V and Pingtou Brother's IP. Secondly, the company is encountering some algorithmic compatibility issues. Pingtou Brother's IP compatibility and continuity are particularly good. If the corresponding algorithm is optimized to the extreme on one chip, the company will have much less workload when making the second model.

In terms of AI, Li Xiaohan revealed that Yunzhisheng is currently working on the medical version of ChatGPT and plans to launch it in the first half of this year. It believes that the main difficulties currently facing are uncertainty, with data and algorithms available, but whether the final results meet customer expectations still needs to be continuously explored.

From the earliest years in 2012 and 2013, deep learning began to be widely applied at the AI level, to AlphaGo in 2016, and then to the recent ChatGPT, the climax wave after wave. Now, people's doubts about this matter are gradually dispelling, and it will indeed affect all aspects of us. Especially for ChatGPT, for AI companies, he metaphorically refers to it as a "nuclear bomb". In the past, everyone used to fight against planes and tanks, but now there is someone with a nuclear bomb in their hand, and you have already lost before you start.

Li Xiaohan stated that ChatGPT is a significant benefit for the server side, and the demand for computing power has suddenly become particularly strong. The edge side has not yet seen a particularly significant impact, but with the development of technology, such as news reports, the reasoning power of ChatGPT has decreased by 90%. How much will it decrease next? Can it be placed directly on the edge side? If it does happen, it will have a huge impact on the computing power, IP, and ecology of the edge side.