Huayu for engineering students

Chapter 132 Make a "Super Calculator"

"Second son, why didn't you tear off the tape, it's done?" Dadong looked curiously at Wang Xu, who was leisurely basking in the sun by the window.

"Well, it's done, the results have been checked and accepted, just wait for the results. How long have you not paid attention to me? It's been a week since I stopped tearing the tape. Come on, what's the matter, you stick your ass. "Wang Xu curled his lips, Dadong must have something to do with being so idle.

"Stop, stop, eat, watch your words." Zhou Feng immediately stopped Wang Xu's nonsense.

"I just want to ask you if there are any projects you can do there. Isn't this a sophomore? The professional courses have already started, and optics is boring. I see you tossing about this and that all day long, and it turns out that the material department is cheap, can you?" We can’t get some projects for the class.” Dadong was the group secretary of the class last year, but he has resigned this year, but he is still very enthusiastic about the affairs of the class.

By the way, Mizuki's habit is that the branch secretary of each class is changed every year, and the student union is also very mobile, and it is rare for one person to do it all the time.After all, they are all geniuses from all over the world. Who can be better than the other? If you don't convince the public, you have to balance it out.

"Don't tell me, there really are." Wang Xu suddenly thought of an idea.

"Tell me." When the other dormitory brothers heard that there was a project, they all came over one after another, standing or sitting around with lunch boxes.

What, why do you go back to the dormitory for dinner instead of coming back after eating in the cafeteria?Well, because there are no seats at all in the Nine Canteens right now.People can only eat standing up, and there is still no table to put lunch boxes on, so it is better to go back to the dormitory to eat.

In the past few years, Shuimu has opened several canteens in a row, and the serial numbers are all up to sixteen. The new canteens have bright and clean windows, but they are far away, and lazy people don't want to run.

"Let's make a supercomputing game. Although this thing is more from the Department of Electronics and Computer Science. But who are we, automation, we have something to do with everyone, so we can do it if we want." Wang Xu said.

"Pfft~", okay, these brothers, the one who sprayed the rice, the one who sprayed the water, all of a sudden the table was in a mess.

"As for it, it's disgusting." Wang Xu complained about the unreliability of these roommates while wiping the rice grains sprayed on his body.

"You are really capable, you really dare to say anything." Dadong was also wiping the sprayed water, and said while wiping.

"Super calculation, the country has invested more than [-] million yuan and it hasn't been done yet. As soon as you talk about it, you plan to do super calculation. Although you have made some inventions, don't be so outrageous, okay? I know, you can really blow it." Boss Zhan Wenbo obviously thought that Wang Xu was talking nonsense.

Indeed, since the Galaxy billionth machine in 83, it has been many years since the progress of the domestic supercomputer.The term supercomputing was translated from foreign news in recent years, and it was always called a supercomputer before.

Supercomputing is naturally one of them, and it is a particularly clear field. Wang Xu thinks about doing this, mainly because the current computers are really not enough.

But today's supercomputers are ridiculously expensive abroad, and domestic ones are simply useless.

For example, Galaxy, with 200 million floating-point calculations, is an important scientific research product that breaks the monopoly in China.But in fact, everything inside is soldered by hand. Can you imagine that a computer has 40 million solder joints and [-] flying wires, all of which are soldered by hand?

This thing is actually to make things that should be done in the CPU directly with the circuit.It’s as if Wang Xu of later generations had a younger brother who directly soldered an 8086 CPU on a table with thousands of transistors. As a graduation project, the Galaxy is actually something similar.

This kind of thing is not only impossible to mass-produce and batch-manufacture, but even has little use value, because its maintenance is extremely difficult, and any problem with any flying wire or solder joint will take an inestimable amount of time to troubleshoot.

Therefore, the most important role of the Galaxy is not to use it for anything, but to tell the world that we have it, and then the international community will lower some conditions for purchasing supercomputers and break some blockades.

Didn't you see that as soon as the Galaxy was released, several supercomputers were successfully introduced into China. Although there are still many restrictions including that they can only be used for specific purposes and must be used by foreign personnel, at least they can be bought.

Moreover, the current supercomputer cannot perform the operations that Wang Xu needs. He does not have a general-purpose operating system or commercial software, and all calculations are completed by existing programs.

Therefore, Wang Xu came up with the idea of ​​making a simple version of supercalculation.But this thing, Wang Xu can't make it alone. Although there are selling technologies in the system space, after all, computer special effects, and things like VR, but they account for an important proportion of entertainment.

But Wang Xu can’t afford these gadgets for the time being. Although they are not expensive, they still need to be saved for a while. Of.

But this doesn't mean that Wang Xu can't come up with something now. If one person can't do it, then a group of people can do it. The most important thing Mizuki lacks is talent.What is the purpose of Wang Xu's desperate efforts to build up his reputation? Isn't it to use the resources of Mizuki as much as possible? Is there any sheep with more fur and longer hair than the one in front of him.

Today's personal computers have entered the 32-bit era, but the power of a single computer is still too weak, and minicomputers are too expensive. CPU upgrades are more limited by Moore's Law, and it is not so fast to double.

But now the mainstream supercomputing is still the vector computing mode like Cray, which is a kind of parallel computing. To put it bluntly, this thing is the computing mode of the graphics card of later generations. , such as the Titan machine in Oak Ridge.

The current Cray machine, you can regard its core as a super-sized GPU piled up by a bunch of RISC chips, and these chips are the cores in the GPU.

After adding some input, output, storage and other devices, the performance is really not as fast as a later generation graphics card.

We can almost say that after miniaturization and integration of Cray machine technology, it is the basic GPU.

Moreover, although this thing is fast, it is very expensive, super expensive, and consumes super electricity, which is not suitable for small and medium-sized enterprises at all.

So how can we get a computer with simple raw materials and high computing speed?
In fact, later generations have the answer, and that is distributed computing.

Distributed is a special operating system that uses a special operating system to send detachable calculation steps to different computers for calculation, and then aggregate the calculation results. This thing is actually the predecessor of this cloud and that cloud that everyone is familiar with.

In fact, parallel and distributed are consistent in basic principles, except that they are not distributed to external computers, but to different internal CPUs, so the speed is faster, but the design is more complicated, and the number of CPUs that can be integrated is also smaller. limited.

However, talking about speed regardless of the application is playing hooligans. First of all, you have to figure out what you are going to do.

If you want to be efficient in parallel algorithms, first of all, what you need to do must support parallel operations, that is, the original operations need to be disassembled into a large number of irrelevant simple operations.

For example, 1+2+3+4, you can easily split it into two irrelevant operations of 1+2 and 3+4, and then calculate the sum of the results. It takes 3 clock cycles to calculate on a computer , while computing on a parallel computer with two CPUs, only two are needed, which increases the speed.

However, some things cannot be split in this way because subsequent operations require the results of the previous step, so parallel operations cannot significantly improve efficiency.For example, 1+2*3, no matter what kind of machine it is on, it takes two clock cycles to complete the calculation.

Distributed also has similar problems. If your work can be divided into many similar subtasks, then distributed is ideal. For example, animation rendering can send materials to many computers to render each frame separately.

After the rendering results are obtained, they are synthesized first, then divided into segments, and sent to multiple machines for video generation.By repeating this, you can get a complete rendering animation.

Audio processing is actually similar to video processing. Audio can also be segmented and processed separately. Different audio tracks are originally different subtasks, so audio processing can actually be used processed in a distributed manner.

It's just that audio processing is much simpler than video, and there is no need to split it so troublesomely. A DSP in the future can handle it directly.

But the current various CPUs and DSPs are not so powerful, so it is a little more troublesome, and some distributed or parallel technologies need to be used.

A high-performance PC is used as the host machine, responsible for human-computer interaction and task distribution, and multiple sets of computers equipped with multiple sound cards are used as computing units to process signals.

Equipped with a centralized storage unit, use the RAID method for transit data storage, use it as a cache, and multiple tape drives as permanent storage devices, well, the general structure is like this.

Does it look familiar?That's right, many small workstations for three-bit image rendering in later generations are of this model, but the sound processing unit is replaced by a graphics card.

In general, doing such a system construction now is actually more or less worth the candle, because the performance improvement caused by various restrictions will be extremely limited.Moreover, this thing already exists in foreign countries, and it is quite developed. Otherwise, where do you think the equipment of those top recording studios abroad come from?

But the problem is, this thing is too powerful for signal processing, and it can be applied to, um, you know.Therefore, it was banned and not allowed to be sold, so Wang Xu had to do it himself.

Moreover, the accumulation of this technology is meaningful, and the significance is so great that it can continue for decades.

The operating system, time-sharing system, interactive system, network routing, information exchange, etc. included in this are many, many, and the smooth development of the past is the cloud of the future, and there is no need to change paths.

Wang Xu intends to use a little of his initial capital, and then squeeze the big sheep Mizuki to get out the super-simplified basic audio version of Future Cloud.

When Wang Xu explained the detailed needs to several roommates, he saw their contemptuous eyes.

"What kind of eyes do you have?" Wang Xu said angrily.

"You call this super calculation?" Boss Zhan said angrily, obviously he was feeling sorry for the disgusting lunch just now.

"Super audio computer, supercomputing for short, is there a problem?" Wang Xu poked his neck.

"Get out." Several roommates raised their feet to kick at the same time.

A few people were fighting for a while, and college roommates are like this. Everyone thinks they are dads. Well, at least Wang Er thinks so, and they probably think so too. It seems that this tradition has not changed in later generations.

"Okay, let's stop making trouble, just say what you want to do."

"It can be done, but it is estimated that only a few of us can't do it. Your plan has too much content, and there is even an operating system, which is fatal. This is something that a few students can do." This is Dadong's complaint.

"Yeah, do you really think we can do this?" Boss Zhan was a little unconfident. This guy usually pushes the boss like his second child, but he is still cowardly when things happen.

"Is there anything I can't do? If I don't have time, I can handle this operating system by myself. I guess it will take a month or so." Wang Xu raised his chin in embarrassment.

This is really not bragging, there are source codes for this in his skill infusion, um, it is the equipment maintenance technology.Just knocking out so much source code will kill you.Moreover, the source code is only applicable to special equipment, and it is impossible to copy it, so if you want to realize it, you have to rely on people to pile it up.

"Just brag, you have to come up with a plan, and you can't just say it. If the plan is feasible, you can go to Uncle Qiang and report it to the top. I think that with your current favor in the department, it is best to apply for a plan." The project should be fine.

But with so many people involved, it may be necessary to cross departments. It is best to involve the Youth League Committee and the Student Union. Their organizational and coordination skills can still be of great help. "Even though Dadong is young, his vision is not low.

Back then, he was a provincial-level three-good student, and he was recommended by Mizuki.Then he worked as the secretary of the Communist Youth League for another year. If he hadn't been lazy, he should be able to join the student union of the department by now.

Oh, by the way, Uncle Qiang is the current secretary of the Communist Youth League, named Wu Guoqiang. Because he is quite trusted by everyone, everyone likes to call him Secretary of the Communist Party of China, or Uncle Qiang for short.

"Well, that's fine. I'll make a plan right now, and I'll toss about it quickly. Recently, I've only been tossing about the materials department, so I can't let the people in my hometown be idle." Wang Xu raised his head and sighed, and was laughed and scolded again.

In the next few days, Wang Xu spent his time on this supercomputer project. Although it is a bit exaggerated to call it a supercomputer, others don’t know. Doesn’t Wang Xu understand it? After the technology in it is accumulated, it is really Super calculation model ah.

-

(Chapter 1, [-] words, I am too lazy to divide it into chapters [-] and [-], please read the longer one if you are troubled)
(End of this chapter)