September 15, 1994.

Shencheng Airport.

Pang Xuelin and Liu Qi stood together at the passenger exit, watching Ni Guangnan and an old lady in her sixties walk out with more than a dozen men and women of different ages.

The old lady was dressed elegantly, although she was in her sixties, her skin was fair, and her face was exquisite, but she could vaguely see the demeanor of her youth between her eyebrows and eyes.

Pang Xuelin took the initiative to greet him: "Mr. Ni, welcome to Shencheng, who is this?" ”

Pang Xuelin's gaze turned to the old lady beside Ni Guangnan.

Ni Guangnan said with a smile, "Mr. Pang, let me introduce to you, this is Academician Xia Peisu of the Chinese Academy of Sciences, I heard that you are going to engage in Hisilicon Semiconductor, she came to take a look this time." ”

"Academician Xia, hello!"

Pang Xuelin shook hands with Xia Peisu and smiled.

In fact, long before Ni Guangnan made an introduction, Pang Xuelin vaguely guessed the identity of the other party.

Xia Peisu, academician of the Chinese Academy of Sciences, the founder of China's computer industry, known as the mother of China's electronic computers, is an extremely rare female computer expert in this era.

In 1958, she was responsible for the design and development of the general-purpose electronic digital computer 107, which was China's first self-developed general-purpose electronic digital computer, which proved with facts and actions that the Chinese had the ability and ambition to design and develop their own computers.

She proposed the principle of the maximum time difference pipeline that greatly increases the computing speed of the computer, which greatly shortens the clock cycle of the pipeline computer.

In addition, she effectively solves the waveform distortion problem that occurs during high-speed signal transmission in mainframe computers.

The research results have been adopted or referred to by many domestic units that develop large-scale high-speed computers. Based on this, she also formulated specific engineering design specifications for a number of high-performance computers that she was responsible for designing and developing, and achieved good signal transmission results.

She was responsible for the successful development of the high-speed array processor 150-AP, the computing speed of the 150 computer is 1 million times/second, and the computing speed of the 150-AP has reached 14 million times/second, which is higher than the computing speed of similar products embargoed by the United States to China at that time, which has received great attention in the world and made a significant contribution to China's petroleum exploration.

She presided over functional distributed computer systems and led the team to successfully develop the GF10 series of computers, which later became the focus of research in the Institute of Computing Technology for a period of time.

In addition, she also founded the Journal of Computer Science, the most influential journal in the field of computer science in China, and the journal of computer science and technology, which is distributed abroad. He presided over the compilation of the "English-Chinese Computer Dictionary".

In later generations, Academician Xia Peisu was known as the mother of Chinese computers and had a lofty position in the field of computers.

"Mr. Pang, hello!"

Xia Peisu's gaze looked at Pang Xuelin with some scrutiny.

She had heard a lot of rumors about this young man, what a genius boy, the youngest richest man, a young entrepreneur.

But what attracted Xia Peisu the most was the paper written by Pang Xuelin on the outline of computer and Internet development.

The paper's high-level description of the entire ICT industry gives people a dreamlike feeling of breaking through time and space and seeing through the fog of the future.

Because of this, Xia Peisu became very interested in Pang Xuelin.

It's a pity that when she wanted to talk to Pang Xuelin, she learned that Pang Xuelin had gone to the United States.

Nearly half a year has passed.

This time, she unexpectedly learned from Ni Guangnan that Pang Xuelin would establish HiSilicon Semiconductor in Shencheng, and invited Ni Guangnan and his team to settle in.

Pang Xuelin and Xia Peisu exchanged a few words, and then, Ni Guangnan first introduced Pang Xuelin to the team members he brought out from Lenovo.

These people are all talents in the field of microelectronics that Ni Guangnan found a way to dig from the Chinese Academy of Sciences and various universities, and most of them are not obvious, but Pang Xuelin did not expect to find a few future bigwigs from these people.

After all, this era is a dark era for China's semiconductor industry, if it does not intervene by itself, China will at least have to explore for many years, and after many setbacks, it will slowly get on the right track after 2010.

It was the two students brought by Xia Peisu who attracted Pang Xuelin's attention.

One of them is Hu Weiwu, the father of Loongson in the later generations.

This is a bull man.

Hu Weiwu was born in Yongkang, Zhejiang Province, to a family of teachers. When he was born, his body was relatively weak, and his father named him "Weiwu", expecting his son to be strong, burly, and heroic.

In September 1986, Hu Weiwu entered the University of Science and Technology of China as the top student in the college entrance examination in Yongkang County, Zhejiang.

The University of Science and Technology of that era was still above the Qingbei Dynasty.

Of the 6,700 or 700 students enrolled at HKUST, 13 are top students in the provincial college entrance examination.

This shows Hu Weiwu's strength in high school.

After entering college, Hu Weiwu showed amazing talent, and in college, he used 74xx chips to weld an 8086 CPU.

In July 1991, after graduating from the University of Science and Technology of China, Hu Weiwu entered the Institute of Computer Technology of the Chinese Academy of Sciences to study for a doctorate without examination.

During his Ph.D., his mentor was Academician Xia Peisu.

From undergraduate to master's and doctorate, Hu Weiwu has always been in China, which can be called a real "local faction", and he even jokingly called himself a "turtle faction".

His doctoral dissertation has won the "Special Award for the President of the Chinese Academy of Sciences" and the first "National 100 Outstanding Doctoral Dissertations" award, one of which is to go abroad at public expense.

But Hu Weiwu still decided to stay in China and willingly became a "turtle faction".

There is no doubt about Hu Weiwu's level, and the two textbooks he wrote, "Fundamentals of Computer Architecture" and "Computer Architecture," can be called classics.

When Pang Xuelin was studying computer-related materials in the real world, he read the textbooks written by Hu Weiwu.

It's a pity that fate dictates.

At the beginning of 2000, when the Institute of Computer Science of the Chinese Academy of Sciences decided to develop a general-purpose processor, Hu Weiwu chose the MIPS architecture instead of the ARM architecture, which later flourished.

But it's also hindsight.

In fact, ARM did not develop well before 2000, when the strongest core design that ARM could come up with was just the very original ARM11, with no out-of-order execution, no multiple launches, and no caching system.

Loongson's original goal was desktop and supercomputing systems, requiring high computing performance to compete with the X86 system.

In contrast, the MIPS architecture in the 90s has achieved out-of-order four-launch, deep pipeline, branch prediction, register renaming, load-store speculation and other advanced designs, and is one of the CPU architectures that can compete with the X86 system in the world, and it was once brilliant.

In the 90s, home game consoles that swept the world generally used MIPS architecture CPUs, such as the well-known R3000A in SONY's PLAYSTATION1, the later R4000 of PSP, and the R4300i of Nintendo N64, as well as Motorola, Ericsson, Nokia, Cisco, Qualcomm, SUN, 3COM and other communication giants in switches, routers, and communication equipment have used MIPS chips.

In the case that the patent barriers of the x86 camp could not be broken for many years, the options that could start for China's CPU at that time were nothing more than MIPS, ALHPA, ARM.

ARM is still embarrassed in its infancy, and the MIPS architecture is used by a large number of Japanese technology companies as the main CPU of user terminal equipment for its excellent computing performance, which has important reference value, so Hu Weiwu, the father of Loongson, chose MIPS.

It's just that no one expected that the MIPS architecture, which was still hot before, would fall apart in the end, and Loongson would become the only user of this architecture.

Loongson began to operate independently early on, and was self-financed.

In the nearly 20 years since its birth, Hu Weiwu's team has only relied on meager profits and the strong support of the Chinese Academy of Sciences to support the development of our country in the fields of supercomputing, military, and aerospace as much as possible.

And a set of independent chip system has been formed.

The total R&D expenses in the past 20 years are only less than 2 billion, which is equivalent to a fraction of Intel's annual R&D expenses.

In the absence of full support from the state and the support of related industries, Hu Weiwu can be said to have achieved the ultimate in what he can do.

In his 20 years of teaching and scientific research career, Hu Weiwu has also brought out many top talents in the field of microelectronics.

Among them, the most famous is Chen Yunji, who later founded the artificial intelligence chip Cambrian.

In later generations, this bigwig led the Intelligent Processor Research Center of the Chinese Academy of Sciences to develop the world's first deep learning special processor chip, which can be regarded as a pioneer and pioneer in the field of artificial intelligence chips.

"Mr. Pang, what's wrong?"

Xia Peisu couldn't help but feel a little strange when he saw that Pang Xuelin's eyes were always staring at his disciple.

Pang Xuelin shook his head and said with a smile: "It's nothing. ”

According to the calculation of time, this year should be the last school year for Hu Weiwu to study with Xia Peisu, such a big man with a solid foundation and unlimited potential, if he does not fool him into working under him, Pang Xuelin feels sorry for himself.

Next, Pang Xuelin and Liu Qi took Xia Peisu and Ni Guangnan on a bus, and first led everyone to settle down in a five-star hotel in Shencheng.

After lunch, after picking up the wind and washing the dust, Pang Xuelin took Ni Guangnan and Xia Peisu to the Shenyi Industrial Building in Nanshan District.

It is also the headquarters of Huawei.

Pang Xuelin has already spent 500 million RMB to buy this building completely, ready to be used as the headquarters of Huawei and HiSilicon in the next few years.

Warwick also saves from moving around as the scale expands.

After arriving at the Shenyi Industrial Building, Ren Zhengfei and the others had been waiting for a long time.

Pang Xuelin first asked Ren Zhengfei to pull the people brought by Ni Guangnan over to visit, and he himself took Xia Peisu and Ni Guangnan to take a walk around the building to familiarize them with the next office environment, and then Pang Xuelin took them to the conference room of Huawei Company.

Ni Guangnan already knew at this time that Pang Xuelin actually owned 49% of Huawei's shares, and couldn't help but sigh with emotion: "Mr. Pang, Huawei's current program-controlled switches are selling very well in the market, and the market-occupied by the program-controlled switches developed by Lenovo is almost eaten up by you, but I didn't expect Huawei to be your company." Coupled with the Transwarp Technology, which is making money every day, I think you probably deserve the title of the richest man in China. ”

Pang Xuelin smiled and said: "Ni Gong, these are just false names, to be honest, making money doesn't mean much to me, what I value more is whether I can convert money into the scientific research resources we need." ”

Ni Guangnan nodded and said, "Mr. Pang, then I won't talk nonsense with you, if HiSilicon Semiconductor wants to develop chips next, have you considered the architecture?" At present, there are not many people on the market that can compete with the X86 architecture. ”

Pang Xuelin shook his head with a smile and said, "Ni Gong, I originally wanted to buy the ARM architecture and do it myself, but it didn't work out, and the British refused to sell ARM to us. So after thinking about it, I think that if we really make a chip that really belongs to us, then we must design a set of our own architecture system. ”

"Our own architecture?"

Ni Guangnan and Xia Peisu glanced at each other, and said in surprise: "Mr. Pang, this difficulty is not as simple as you think. ”

The most difficult thing to develop a new architecture is not the technical level, but the time, money, and ecosystem.

On the one hand, the R&D structure actually requires time accumulation and a large amount of financial support.

Without these two things, there is basically no way to talk about the development of a new architecture.

But at the same time, the ecosystem is also a top priority.

X86 desktop processor market, Arm low-power market, MIPS cheap market.

Even if you develop a new instruction set, but others don't use it, you don't have the money to take it, and you can't continue to develop it if you don't have the money, it's an endless cycle.

Therefore, in a sense, the importance of the ecosystem is far greater than that of technology, and it is difficult to overturn once the advantage is established.

No matter how smooth WP is, it is like cannon fodder in front of Android.

How many applications are willing to optimize the new instruction set for you? And how many people can identify with and use it?

It's all a problem.

And the instruction set needs the support of the system, you pirate other people's Windows system, Linux system, is it useful to say that it is developed by yourself?

After all, you can't develop your own instruction set on other people's systems, and if you want to develop your own good instruction set, you must first develop a good system and a good CPU.

Therefore, in a sense, the importance of the ecosystem is far greater than that of technology, and it is difficult to overturn once the advantage is established, just like the MAC is forced to be high, and it is not smooth to do it, and it is also cannon fodder in front of Android.

How many applications are willing to optimize the new instruction set for you? And how many people can identify with and use it?

It's all a problem.

And the instruction set needs the support of the system, you pirate other people's Windows system, Linux system, is it useful to say that it is developed by yourself?

After all, you can't develop your own instruction set on other people's systems, and if you want to develop your own good instruction set, you must first develop a good system and a good CPU.