Chapter 281 Developing a Super Compiler
,
Although Poincare and Turing often quarreled, they were very serious when it came to academic matters.
Turing was not the kind of person who pretended to know everything when he did not. If he had a real question, he would ask until he understood it.
Lu Shan agreed with Turing's plan and got involved. This is the focus of his recent work.
According to the new perspective provided by the Riemann hypothesis, the computer field should benefit from encryption algorithms and cryptography, since a large part of this field is about prime numbers.
Unlike the RSA encryption algorithm, it relies on the difficulty of large number factorization to ensure security.
The Riemann hypothesis has a high-level summary of the distribution of prime numbers. If you master it, it will be easy for you to crack it while others cannot crack it for you.
Therefore, the first requirement of Lu Shan and Turing's compiler is security, but the Riemann hypothesis has expanded human thinking far beyond imagination, and has provided more angles for exploring the world.
This is called analogy. Many internal things in science disciplines are consistent. Ultimately, it all comes down to the most basic mathematical logic.
With Lu Shan's joining, the compiler quickly took shape, the framework was quickly built, and the details were constantly filled in.
After taking a breath, Lu Shan rubbed his sore eyes and took a break. Turing then checked the work he had done before to see if there was anything missing.
After several days of intensive work (in Singularity Lab time), the intelligent compiler was finally implemented.
The intelligent compiler developed by Turing and Lu Shan has three most important features.
The first characteristic is naturally conciseness. Turing and Lu Shan both hate lengthy codes. Not only are they painful to type, but they are even more painful to maintain in the future.
Do you know why many games and systems are getting bigger and bigger?
That's because the code was not well organized at the beginning, and it was impossible to optimize and simplify it later.
Each upgrade involves adding code to the existing code. The code gets higher and higher, and the shape gets more and more outrageous. People in the industry call this a "code shit mountain."
To put it in a more vivid way, it's like a pile of garbage, stacked very high and crooked, but it can still run.
It is a rule in the industry that if the code works, don't change it, so you must be concise from the beginning and keep the code as small as possible to make it easier to maintain in the future.
The second important feature is stability and scalability.
Stability is very, very important. As the mountain of code piles higher and higher, the hidden dangers of stability are very great. It is very likely that a small mistake will cause the entire mountain of code to collapse.
But if you want to optimize the past code and make the system more concise, it is equivalent to taking out the original things and stuffing them back in, and then the program can still run. The difficulty of this is beyond description. You can imagine it yourself.
If it is square from the beginning, it will be much easier to maintain in the future. The smart compiler has such a feature, it is standardized.
The same things will be understood first, and then implemented with the same code to avoid overly lengthy descriptions.
Modular production, standardized process, plug and play at any time, this is how the intelligent mutator manages the code.
In the future, the previous mountain of shit code can be optimized without system crashes. Future systems may even use less code and have lower CPU requirements.
Extensibility actually refers to the learning ability of the intelligent compiler.
When accepting user compiled code, the intelligent compiler will learn human language and logic. As more and more application software is compiled, the compiler will become smarter and smarter, similar to the function library of traditional compilers.
The same situations will be recorded, and similar situations will be adjusted according to the variables. In short, it increases your own knowledge base.
What are the benefits of this?
Originally, code was typed line by line, and the same things had to be repeated again and again.
The intelligent compiler directly puts what it has seen in the past into your program, and if it is not right, you can just modify some things.
The underlying logic is still to convert into 0 and 1 for the machine to understand, but it used to be manual and is now automated, saving considerable time.
To give a simple example, when the compiler is just completed, the user wants to calculate the area of a rectangle.
You tell the compiler that the operation is length times width.
Keep using it. When the compiler sees that you want to know the area of the rectangle, it will automatically multiply the length by the width, and there is no need to express the logic.
If you are editing a game, you can save a lot of time. Actions, battles, and graphics are all computationally intensive and highly repetitive processes. With this, you can really free up a lot of manpower and shorten the construction period.
In the future, humans will be able to create more unique games and give full play to their imagination.
Moreover, as it is used more often, the latest logic can even be continuously optimized to replace the past logic.
When the compiler finds that the logic describing a certain content has been upgraded and the number of lines of code can be reduced, it will automatically update. This is the learning ability.
In Lu Shan's opinion, this is just the beginning. In the future, the compiler needs to be more intelligent and have the ability to interact with people. This ability is like AI, which requires training and accumulation.
The last feature is the need for advanced hardware support. For the time being, we are still using Western CPUs. There is no other way because the West has them but Xia Country does not.
The Xiaguo Academy of Sciences has to think of a solution to this problem. Maybe they can find a way to buy more advanced CPUs from the West.
The last feature can actually be regarded as a disadvantage of the compiler, as it has relatively high hardware requirements.
The compiler needs to have amazing performance, perform a large number of instant calculations, and store a large number of commands. It cannot be done without powerful hardware.
To be honest, Lu Shan is not very satisfied with foreign CPUs. According to the nature of Westerners, what good things can they sell?
According to Lu Shan’s assumption, he would have to develop his own CPU in the future. By then, the CPU, memory, and motherboard would all be made of high-performance topological semiconductor materials, and only then would the compiler be able to fully play its role.
The CPU problem cannot be solved for the time being, so let’s solve the memory problem first.
So Lu Shan talked to Zhang Kaifeng and ordered a batch of high-performance memory. 1T is enough. It would be a waste of CPU if there is more. After the CPU performance improves, we will make new special memory.
"Chips are always a big problem!" Lu Shan rubbed his forehead. This thing is already of national strategic level. The level of chips determines the level of development. Lu Shan doesn't think much of foreign CPUs. If he develops topological semi-metallic semiconductor materials, it is destined that everyone's chip manufacturing path will be completely different in the future, and the performance of his own materials will definitely be higher.
But... I still have to use theirs for now.
After receiving Lu Shan's request, the Xiaguo Academy of Sciences immediately began to deal with it. It doesn't matter what they do specifically, as long as Lu Shan has the right way.
Furthermore, if Lu Shan comes up with something good, it will definitely be very beneficial to Xia State, so we must pay attention to it and do it as soon as possible.
However, Lu Shan's requirements this time were rather vague and difficult to achieve. He said he wanted the best equipment possible, but did not specify the model or performance.
So the Xiaguo Academy of Sciences put in a lot of effort and tried their best to search for it, and finally completed the task.
When they looked at the most advanced equipment they could buy, they hesitated. They had to give Lu Shan something they had spent so much effort to buy. It would be a great help if they used it for themselves.
"Look at what you have accomplished!" Xia Guangming was speechless. He could understand everyone's feelings, but it was obvious that this approach was not advisable.
"Once we promise to do something, we should complete it. This is called keeping our word." Everyone seemed a little skeptical.
Xia Guangming simply gave them a heart attack: "Do you think this thing is more valuable in your hands, or in Lu Shan's hands?"
"Uh..." This question really hurts the heart. Who dares to compete with Lu Shan? At the beginning, with almost no resources, he only relied on his brain and a low-performance computer to run the model to come up with topological semimetal semiconductors. It's like a person using a pile of scrap copper and iron to build a super awesome fighter jet that can fly.
"If no one dares to guarantee it, then send the things to Lu Shan and don't delay his business." Xia Guangming waved his hand, and his men obediently delivered the equipment over.
The things were delivered to the warehouse agreed in advance. Lu Shan went in when no one was there, spent points to bring the things into the Singularity Laboratory, and then the warehouse was empty.
With the high-performance equipment, Lu Shan did not rush to put it into use. Instead, he followed the convention and made modifications according to his own ideas and needs.
"What a crappy combination! What a crappy system! Such a good thing is wasted!" Poincare also lent a hand, complaining as he worked on it.
Good hardware is good, but when put together, it's really unwatchable.
"It feels like people nowadays are trying their best to push up the performance data of every part, but they don't pay attention to debugging." Turing also kept shaking his head.
Lu Shan said sullenly: "Even so, they are still better than Xia Country's products..."
Can you understand the logic of this matter?
The hardware of Fujiwara Takumi's AE86 is so poor that it was adjusted to the limit by his father, but no matter how hard he tried, it could not outrun those good sports cars.
Trying to maximize performance by being stingy is a sign of poverty. Others can be better than you by just stacking and wasting things randomly, so why bother?
It's not all like this in the West. They will rack their brains in the ultra-high-end fields, but when it comes to commercial things, they just let it go because there are no rivals anyway.
This is where Lu Shan is unhappy. Sooner or later, he will let them experience this feeling.
The three of them worked overtime and cooperated in the Singularity Lab, and the compiler gradually became more complete. They conducted small and large tests continuously, and the results were very good.
According to Lu Shan's idea, this compiler must be fool-proof, that is, anyone with little programming experience can use it! Many people have good ideas, but they don't have the ability to program, so their good ideas are wasted.
Therefore, we must make the compiler intelligent to help more people put their ideas into practice.
The interface design is very, very simple, and looks like Word. You enter requirements or directly encode on the top, and record them on the bottom. It can also run synchronously, and there will be a reminder if it is inappropriate.
You can ignore it or modify it according to the prompts.
If the user can eventually round off the logic, or even get a better logic, the compiler will automatically record the update.
The main idea is that you and I learn from each other and improve together.
When inputting compilation requirements, you can use Chinese and English, and you can also use mathematics and mathematical symbols to describe and compile.
This is a feature that no compiler on the market currently has. Has anyone ever seen a compiler that can do what I want to achieve using plain language?
Just type the code honestly.
This is the core competitiveness.
The compiler's database records all mathematical logic and calculation methods, human logical language, especially Chinese language, and the corresponding logical methods of mathematics.
And this function is still being improved.
Among the students trained in the natural science laboratory, some are majoring in this category.
Even with the development of compilers, it may eventually be able to understand philosophical issues and promptly absorb new Internet vocabulary, new era words, events or logic.
This information will be communicated by talents in related majors with talents in mathematics, and cooperate with computer professionals to convert human thinking into one-to-one mathematical logic, and then stored in the compiler's database.
"Do we need to record all new words on the Internet? There is no need to do that. After all, they are not the basis of language, but the product of evolution and development."
Turing hesitated, "Besides, many new Internet words are just a passing fad and don't stick around."
Lu Shan shook his head and said, "You can't think like that. You don't need to care about those boring and illogical new words on the Internet. But many new words are actually very interesting. It would be inappropriate to discard them all. Language itself is a process of continuous development. The writing of the Xia Kingdom was not formed in one day. Many words did not exist in ancient times. As long as they can accurately express the meaning and have a certain level, they can be collected.
Just imagine that when a person is old but still has new ideas, he must use the speaking habits he had when he was young.
If we don’t record it, they won’t be able to learn new vocabulary and it could be a waste of everyone’s intelligence.”
"That makes sense. The English language today is indeed not what it used to be. If we don't record it, the language will be missing out on the evolution process."
Turing finally understood Lu Shan's intention, but this was a long and complicated task. Screening out bad new words was a huge task. No wonder Lu Shan recruited students with majors in Chinese language and literature, English, and philosophy, and planned to set up a team to do this job.
This compiler is now available and is very, very advanced. Even if it is not upgraded, it will be ahead of its peers for a long time.
Lu Shan conducted many tests, such as directly pasting and copying the final math question of a certain province's college entrance examination and asking the compiler to do it, and requiring a detailed process.
The compiler quickly gave the correct steps and answers, which went beyond the function of a compiler and became an intelligent search engine that could also record.
If you switch to another language, you have to input the logic of the mathematical formula bit by bit, teach the computer how it works, and then input the other conditions in the question and let the program run bit by bit.
(End of this chapter)
