Tech Hammer

The phone kept vibrating next to her, but Ning Wei seemed not to notice it.

Recently, he has almost completely given up on turbulence algorithms and devoted himself wholeheartedly to EDA software.

If it weren't for the turbulence algorithm, there would be no problem. In fact, the reason why this project has been undergoing internal testing in the laboratory and has not yet contacted the outside world for large-scale public testing is because it encountered technical bottlenecks.

Currently, the erroneous judgment rate of the turbulence algorithm in the laboratory is stable at about 1.8 parts per 100,000 and cannot be reduced further.

The error rate of 1.8% per 100,000 does not sound high, but if applied to the servers of ticket sales websites and APPs such as 12306, it is very scary.

The user base is too large.

For every 100 million people who use 12306 to book tickets, more than 1,800 people may be misjudged as malicious crawler programs and punished. This is a very bad thing. Not to mention that more than 100 million people use the 12306 ticket booking system every year?

It’s more than a long vacation!

Ning Wei and the three auxiliary researchers did not think of ways. Several people brainstormed together and went through two iterations of the algorithm, but they were still unable to reduce the error rate in the laboratory test.

This makes Ning Wei feel that further breakthroughs in the turbulence algorithm may require him to further deepen his understanding of the entire system. This cannot be rushed. He may simply hang in the laboratory for a while and do internal testing for a while. There is enough data to analyze, and maybe we can find some clues from the data.

So Ning Wei no longer pays attention to the turbulence algorithm. After all, no one in the EDA software project team knows the structure and difficulties of this software better than him.

The biggest difficulty is actually to simplify the chip design.

Give a simple example.

When using mainstream EDA software for back-end design, the first step is usually to set up the standard library.

The required data can be roughly divided into three categories, namely logical layer, physical layer and design.

These three categories can be subdivided into six types of files, including process files, unit libraries, sub-libraries, netlist files after synthesis, TDF files that constrain Pad positions, and timing constraint files given after synthesis.

Then there is the overall layout of the chip, which is an important step that directly affects the chip area, speed, signal integrity and design cycle. Corresponding positions must be assigned to modules, input and output interfaces, power pads, etc.

Next is clock tree synthesis, which is also a big challenge. Think about it, after gathering tens of millions or even hundreds of millions of transistors, it is not easy to transmit the clock signal to every local area of ​​the chip with zero deviation.

Next is the overall wiring, which is also the most important physical practice process of the entire chip design.

This is followed by a design rule check.

After all these are done and there are no problems, let’s enter the simulation process...

According to Ning Wei's idea, the above steps should be completely reduced to one step, and the first-level chip architecture that completes a specific type of task should be integrated into the object library, which is the container, and then the second-level architecture is the group, and then proceed downwards. Divide...

The operator only needs to make simple combinations, and after simple repairs and deletions of the existing structures, he can directly enter the simulation process.

This is why this EDA software is purely for designing various AI chips.

If we want to design a general-purpose chip in this fool-proof way, even if Ning Wei really loses all his hair, it is impossible to solve the problem of fool-proof operation.

But it can be done in the field of AI chips.

As far as the current situation in the field of AI is concerned, the so-called intelligent algorithm is the process of quickly seeking the optimal solution. These algorithms solve complex engineering problems by simulating some natural processes.

There is also data-based machine learning, which mainly uses known data to predict unknown data and so on.

So Ningwei’s design concept is very simple.

For example, it first carries a standard convolutional neural network learning container, and then integrates groups of various machine vision categories under the subcategories, such as picture recognition, face recognition, advanced expression recognition, gesture recognition, etc.

A group is set up under the group to complete specific single functions such as weighting, summation, activation, transfer, and regression.

The operator only needs to enter the type of chip he wants to design, and the container frame in the object library will be jumped out first, and then the operator can choose the functions to be implemented under this frame.

It is difficult for software designers to make it easy for operators.

Ningwei's goal is to design at least five container structures in the object library before the launch of this EDA, as well as corresponding fusion group structures.

Once it is launched on the market, it will be simple. If someone uses it, the high-level design team can design the container, the supporting groups, groups and components by themselves after learning the various standards and instructions of the software, or use the already defined Components redesigned the container structure and supporting groups and groups with more powerful functions. After uploading, it was included in the object library after multiple verifications by the review department.

But at this stage, he has to do these tasks himself. After all, the professors assigned to the laboratory are good in programming skills, but they are still a little behind in mathematics.

Finally, after three hours of busy work and using the previously defined Yuanhe group to construct a group structure under a container, Ning Wei leaned on his chair and let out a sigh of relief. Working on scientific research projects is really tiring, especially when you encounter difficult problems.

The turbulence algorithm encountered a bottleneck, and the EDA side couldn't do without people. Ning Wei began to reflect on whether he was really too greedy and thought of scientific research problems too simply.

Picking up the phone casually, I realized that the dormitory group was particularly lively today.

I have already read dozens of messages early in the morning.

My mind was a little dull, so I just wanted to see what these people were doing and rest my mind.

Then I saw the complete steps of the classic face change.

At first, several people took turns @him in the group, and the threat was very serious.

"Ning Wawa, if you don't go back to the dormitory, we have decided to expel you from the dormitory!"

"Ning doll, dads miss you."

"Xiao Ning, your wings are stiff, aren't they?"

Probably because he didn't respond, his tone started to soften.

"Little brother, stop making trouble. We have something urgent to discuss. Please reply quickly."

Later, I relaxed completely.

"Brother Ning, Boss Ning, you can't really ignore me."

In the end, they were completely submissive, and some even lost all moral integrity.

"Dad, Dad Ning, please send me a message back!"

It's pretty good. It's hard to describe the psychological satisfaction that comes from being able to see my roommates take the initiative to call me daddy.

So he replied with "?"

Mr. Xuan was the first to react: "Damn it, Ning Wawa, you finally stopped pretending to be dead? What are we going to do with our thesis? You really gave us an idea!"

"I'm telling you, don't cause trouble!" Ning Wei responded with a panda-like expression.

"Dad Ning, please give me some advice on the children's papers!"

"You can teach me!" Ning Wei responded with a panda expression.

Fortunately, this time he didn't whet everyone's appetites and replied directly: "It's almost dinner time. You should first consider which child should be invited to dinner at noon."