Intel's Pentium processor brings two technical concepts, one is superscalar, allowing CPU to execute two instructions in one cycle.

Another one is deep flow, which is to divide an instruction into multiple steps to execute. The flow depth of the Pentium processor reaches level 14. When it develops to the Pentium four processor, the flow depth has reached more than 20 levels.

Super-flow technology has a significant role in improving performance and main frequency, but it is not without its shortcomings.

The biggest drawback is that the ultra-long flow is facing inefficient execution efficiency.

This is also the Ben Four era. The Intel Pentium 4 processor with the same main frequency is often pressed to the ground by the Amd Athon processor with similar frequency.

And hyperthreading can perfectly solve this problem.

In this era when main frequency is paramount, Huaxin Technology's CPU wants to make a name for itself in the market, so it naturally has some considerations on main frequency.

There are only two ways to increase the main frequency: one is to improve the process and the other is to improve the technology.

In terms of process, Huaxin Technology has no choice but to find a way to do it technically.

Wang Anran's solution is to first use Risc thinning instructions on the CPU instruction set, because each instruction of the computer requires additional transistors and electrical components, and most complex instructions can be composed of multiple thinning instructions, and a good balance point is selected in terms of efficiency and performance.

The second is to choose super flow technology. The flow design of Huaxin Technology's first generation CPU is 18 levels and the design frequency is 200mhz, which naturally also has the defect of insufficient assembly line execution efficiency.

The original design of this CPU is no less than that of the Pentium processor in the world.

With the hyperthreading technology, Wang Anran is very looking forward to the simulation results.

But the facts gave him a loud slap.

EDA simulation test results show that Smt hyperthreading technology does not bring significant performance improvements to CPU.

Spending 10% of the number of transistors to build only brings about 10% performance improvement, which is already a useless existence.

There is no need for Huaxin Technology to unveil such an important technology for this 10% performance improvement.

Wang Anran analyzed the reason, it should be that the potential of ilp flow has not been fully discovered at the 18th level, and the CPU design of this era is limited by the number of transistors, and its idle resources are also limited.

Then the question is, where do these 10% of transistors be added?

This is not a problem for a master like Wang Anran. There are many ways to improve CPU performance, and the best ones that apply to this era are the ones.

The final solution is that Wang Anran will add the new CPU to the mature Tage branch prediction algorithm using the new Tage branch prediction algorithm.

The goal of branch prediction is to store it in advance in the cache area of ​​the branch target when a branch processing process occurs.

The essence of branch algorithms is to guess the path of the instruction flow.

So the question is, what if you guess wrong?

The solution to guess wrong is very simple, that is, guess again until you guess correctly.

But the consequences are obvious, the process is stagnant, the CPU execution is inefficient, and the user's experience is that the computer is stuck.

Therefore, an excellent branch prediction method often greatly improves performance in CPU execution efficiency.

...

Fu Yan is a graduate student trained by the Academy of Sciences. He received his Ph.D. in Microelectronics at the University of Cambridge and spent three years at the Texas Instruments European Institute.

This is Huaxin Technology spent a lot of money to dig up turtle talents.

Fu Yan has profound attainments in computer algorithms and has published articles in many heavyweight publications at home and abroad.

I originally thought that when I came to Huaxin Technology, I was too useless, but I didn’t expect to enter the CPU business department as soon as I came in.

Does Huaxin Technology actually have such ambitions?

As a Chinese, Fu Yan, like most of you, has a Chinese core dream.

But can Huaxin make good Chinese cores?

Three days later, Fu Yan changed from suspicion to affirmation and amazement.

From instruction sets to CPU architecture design, Huaxin Technology has an extraordinary foundation.

Is this still a startup company? From the beginning of dozens of people, to now, less than 200 people, can it actually take on the task of CPU design?

And here Fu Yan can also show his expertise in mathematical algorithms.

But today he has seen the methods of his own legendary founder.

The strategy of changing the prediction table is simply changed and a few prediction strategies are added, which increases the accuracy of the entire prediction by 30%.

Wang Anran shook his head. This is his achievements from others. There is nothing to be praised for. The optimization of a prediction table strategy is a concentrated development of dozens of generations of processor technology and the crystallization of the wisdom of tens of thousands of scientific and technological personnel. You should know that the prediction algorithm of the tage branch has won four international awards.

Compared with the original version, the CPU branch prediction algorithm integrated in Huaxin Technology can only be considered a castrated version. There is no way to do this. The original algorithm needs to be fully implemented, and 3 million transistors are not enough.

Wang Anran's low-keyness became an unfathomable meaning in Fu Yan. While it hit his arrogant self-esteem, it also made him calm down and continue to do the following things.

"Professor Fu wants to give branch prediction a better space to implement, and has higher requirements on the width of the branch prediction table and the pre-read data capacity. Under the CPU cache, it is necessary to design another level 2 cache."

Fu Yan shook his head and said, "It's good, but the technology does not work. The second-level cache requires the bus to allocate additional bandwidth, which will affect the communication between CPU and memory.

In addition, sram cache is not only expensive, but also the chip area it takes up will be a problem. We will not have enough transistors to implement other functions."

Wang Anran shook his head. The higher the sophisticated talent, the more likely it is to be restricted by conventional thinking.

Intel is a good industry overlord, but it also happened in the Pentium Three era. In order not to allow the huge SRAM cache to occupy the limited CPU core area, it "creatively" integrates the CPU level 2 cache outside the CPU core, resulting in the CPU level 2 cache that can only run at half speed.

The result is that the CPU is running inefficient and has weak performance. It is at a disadvantage in the K6 architecture CPU of the same period of Amd.

"Professor Fu, we can design a separate bus communication bus for the l2 cache, but it only requires about 2% of the transistors."

“Redesign a bus?”

Wang Anran nodded and continued: "There is no need to use SRAM cache for the second level cache, and the edram cache is also possible. With 128kb of l2 cache, the performance is worth looking forward to."

Fu Yan was surprised. The boss's thinking was so imaginative and unconventional, but he still raised his questions.

You should know that all chip designs today have not yet used edram as a CPU cache case. Is it necessary to be the first company to try the first cpu for Huaxin Technology?

"Mr. Wang, is this edram really OK?"

Of course, although I haven't tried this era, in 9102, edram cache was already a very mature technology.

Wang Anran nodded affirmatively, "If you don't try, how do you know that it's not possible..."

…………

It is undoubtedly the most important step to get started with the EDA design software from the beginning. With the help of EDA software, many difficult problems have become simple.

For example, using EDA software, designers do not need to know how transistors perform logic processing on the underlying layer. They only need to form existing logic gates into logical units and then design these units into larger-scale computing units.

This undoubtedly greatly reduces the design threshold and design difficulty.

The role of the underlying engineer is to face these abstract logic gates and logic units, seek simpler transistors and electrical components, and display them on the physical level.

This is like an assembly line. It used to be a workshop-style design, but now there are two flows, one is a design CPU and the other is a design circuit.

In 9102, this assembly line will be deepened into dozens of design categories, and the professional category of EDA software will also be improved.

Huaxin Technology is now developing in this direction.

The redesigned CPU performed well in EDA simulation tests, which improved by 37% compared to the performance of 200mhz and simple hyperthreading technology.

The number of transistors increased by 16%.

Acceptable range.

Professor Fu pointed to the test results and said, "Mr. Wang, I really didn't expect that with the l2 cache and branch prediction, the CPU performance has improved so much, with nearly 50% performance advantages at the same frequency."

Wang Anran said: "It's just that I can't get to 200mhz, which is a bit regretful."

The CPU frequency is affected by the bus and l2 cache. The heating at the frequency of 200mhz will not be controlled, but at 166mhz, it should be possible to add good heat dissipation. This depends on the specific test situation after the flow sheet is produced.

But even so, it is still very far beyond the frequency of the latest Pentium 586 and 133mhz in the world.

It seems that it is possible to press Intel to the ground and rub...

…………

A month later, Wang Anran returned to the public to realize it, and Wang Anran had already made a decision on Huaxin Technology's CPU plan.

Clark contacted Wang Anran as soon as possible. In Wang Anran's office, Clark expressed his concerns.

"Mr. Wang, can you tell me clearly that Huaxin Technology has decided to enter the CPU field."

Wang Anran nodded.

Clark sighed and said, "It's really disappointing news. Is Mr. Wang trying to use the money from investors to try out unfamiliar projects? I regret your company's CPU project, which is destined to not succeed."

Wang Anran said with great interest: "Mr. Clark, why do you say that?"

Clark didn't even want to explain, but you, Mr. Wang, can't understand the truth that a three-year-old child can see!

If CPU is really so easy to do, why are there only a few companies such as Intel, Amd, Meinida, and IBM in the world?

"Mr. Wang, if your company insists on entering the CPU industry, Goldman Sachs may express doubts about your company's future development, which will have a significant impact on the valuation of Huaxin Technology."
Chapter completed!