With the current chip shortage being a hot topic once again, the blockade of foreign chip supplies and the price surge of domestic MCUs have gradually led to a consensus in the embedded industry regarding the trend of chip localization. Therefore, I would like to discuss the advantages and disadvantages of these two domestic chip architectures in embedded programming.
MIPS has been introduced to China for a long time, and currently, the most successful and widely used ones are Loongson from the Chinese Academy of Sciences and Beijing Junzheng. Loongson has always been a pride of domestic chips, achieving remarkable results in desktop systems, domestic military, and aerospace fields. Junzheng targets the consumer market, producing smart devices such as smart door locks, facial recognition modules, smart doorbells, Alipay and WeChat scanners, wearable devices, etc. Due to advantages such as low cost, high performance, and low power consumption, it has secured a place in the market. However, the market share of MIPS chips is still too small compared to the ARM series, especially Cortex-M, making it appear particularly niche.
As the author has worked on MIPS-related chip development for a long time, I am well aware of the pain points and difficulties involved. The characteristic of embedded development work is that it is generally application-centered. For most embedded engineers, they only need to complete the driver functionality according to the chip datasheet, and then design the upper-level business logic using C language. The differences in architecture mainly involve interrupt handling processes and the usage of MMUs. For general business applications, this does not have a significant impact.
However, for some unknown reasons, domestic chip manufacturers are reluctant to release the chip datasheets, often keeping some information secret. For example, Junzheng’s chips do not focus on creating an ecosystem but rather package solutions for sale, requiring substantial effort for secondary development. The datasheets for Junzheng’s chips are not as comprehensive as those for foreign chips, and it takes an experienced engineer a long time to understand their design and usage. Loongson faces similar issues; although it markets its chips, its software ecosystem and documentation are severely lacking, making the learning curve much steeper. In fact, after working with it for a while, I found that programming practice is not very difficult; the problem lies in the poorly written documentation.
The current issues with MIPS in embedded systems stem from a lack of documentation and usage tutorials, as well as insufficient support for the entire ecosystem. Although Loongson has begun to integrate with domestic operating systems, its upper-layer software ecosystem is still extremely lacking, with industrial auxiliary design, auxiliary manufacturing, and some office entertainment software still being gradually improved. While the differences in architecture do not significantly impact embedded programming, the ecosystem influences whether an engineer is motivated and interested in learning and researching. ARM has excelled in this area.
So, can RISC-V surpass in the domestic chip market? This largely depends on government support and the development of the domestic ecosystem. From the perspective of upper-layer programming, developing RISC-V MCUs and programming ARM chips are fundamentally the same, except for differences in IDE usage; various embedded programming models and philosophies are entirely similar. Lower-level embedded developers mainly focus on making API designs more stable and complete.
Another crucial point is the cost of learning RISC-V. If the architecture requires developers to carry too many conceptual burdens, the positioning of the chip design is inevitably incorrect. A conscientious chip design company should prioritize ease of use and the convenience of enhancing its business logic for embedded engineers. Currently, RISC-V also faces a lack of ecosystem support, but this is different from MIPS; one is just starting, while the other is reluctant to develop. Due to the open-source nature of RISC-V, many people are willing to experiment, research, and learn, and university courses and electronic competitions are keen to choose RISC-V as a topic.
Currently, RISC-V applications are still at the MCU level, but this year will inevitably see more MPU and AI chips emerging. Stacking is a product; with the further development of SiP packaging, the design threshold for chips will lower, and the strengthening of the chip manufacturing industry will follow. Although RISC-V chips are given great hopes, we must face the fact that time is still needed for development.
In the field of embedded application development, architecture is not an absolute barrier. Whether it is Loongson, Junzheng, or RISC-V, embedded programming is common. We should focus on the thinking of programming models, framework design, and business accumulation. By deeply understanding one architecture, the fundamental methods for programming other architectures are also applicable.