Chapter 365: Grab resources from Dream Company!
The speaker is Song Jian.
And such a big boss actually used a humble expression like "asking for advice".
Even before asking a question, he raised his hand specifically.
The noodles are full!
In this case, other people in the conference room naturally returned to silence instantly.
Song Jian reached out and adjusted the black-rimmed glasses on the bridge of his nose:
"Xiao Chang, I probably understand the mesh generation algorithm you just introduced. It requires a large-scale model surface mesh as input, and ensures that the surface mesh of the generated tetrahedral mesh is consistent with the input mesh. Highly consistent.”
Although he is not an expert in the field of numerical calculation, he obviously has a profound theoretical foundation after working in the field of control for many years. It is not difficult to just understand this introduction.
“That's right.”
“So, for larger-scale projects, will this process take up too much memory, causing the program to be unable to run normally?”
Chang Haonan wanted to say a good question at this time.
But considering Song Jian's age and status, he finally held back.
In this era, many so-called "supercomputers" only have 4-8G of memory, which is actually unable to withstand high computing intensity.
Furthermore, it does not mean that a program that requires 8G of memory can be used normally on a computer with 2G of memory by slowing it down four times.
For a specific algorithm, simply reducing the calculation speed will not significantly reduce the memory usage.
But Chang Haonan has naturally considered this issue.
"That's right. It will be mentioned in detail in the document I sent to the research group responsible for this module. When it comes to specific program writing, a multi-level regional decomposition strategy will be used to ensure the parallel efficiency of the method. Specifically, That is to say, the first level of regional decomposition uses planes to divide three-dimensional areas, and the second level uses straight lines to divide two-dimensional areas. Both parts are performed recursively and will be assigned to different processors. "
"Judging from the results of my previous tests, for a supercomputer or server with 8G of memory, provided that the CPU concurrency capability is sufficient, it can handle a grid number of billions of units. As for more complex structures, , you need to wait until the trial version of the software is completed before officially testing it.”
"In addition, for computers with insufficient processors, we will consider providing a downgraded version. Of course, for commercial reasons, this version may be named open version or educational version. Its efficiency is probably that of the full-featured version. It is about one-fifth of the total, but it is still far more than similar products currently on the market such as OpenCASCADE Technology, and it also has some functions that the latter does not have.”
“And when the commercial version of the software is released in the future, we will also selectively open purchase rights according to different customer levels.”
After this clear explanation, Song Jian nodded with satisfaction:
“Okay, I have no questions.”
After two hours of high-density information, most of the people at the scene were already quite exhausted, and the effect would not be good if they continued forcefully, so Chang Haonan directly announced a tea break.
Of course, due to limited conditions, there are no fruits and snacks.
Really only tea is provided.
However, under the current situation, no one is in the mood to eat and drink.
Especially the groups assigned to work on the grid generation module started discussions on the spot with the materials Chang Haonan sent them.
Just like taking classes in college, people will naturally come to Chang Haonan directly.
For example, Yao Mengna and Xu Yang.
The two of them still don't have much psychological baggage.
“Chang Gong, no, Director Chang, Meng Na just put forward an idea. I briefly verified it and I think it seems feasible.”
The latter turned his head in surprise and looked at Xu Yang and Yao Mengna standing side by side.
It seems that the relationship between them has become much better since they came back from Haojing this time...
“Junior...Junior brother, it's like this. When I was at 603, I also participated in a project to improve the turboprop engine's propeller adaptation.”
Chang Haonan thought for a while and felt that what Yao Mengna was talking about should be an improved version of the Y8. However, considering that not all people in the conference room were from the aviation system, he did not specify the specific model.
The original Y8 used a four-blade propeller that was as mediocre and unspectacular as the An-12. By the 1990s, it had completely fallen behind the times in terms of propulsion efficiency and noise control.
So in the original timeline, both the Y8 III platform and the Y9 platform were equipped with updated 6-blade propellers.
Although the current development of Yun 8 has been slightly deviated due to Chang Haonan's intervention at the end of last year, the specific propeller upgrade should not be much different.
“Uh... are you encountering difficulties?”
“That's not true. Although I am not as capable as you two, I can still do this kind of small work.”
Yao Mengna shook her head angrily.
"I suddenly thought after listening to you just mentioned again the idea of converting triangles into pseudo-quadrilaterals. Is it possible to use real quadrilaterals or hexahedral meshes in the process of generating structured meshes? grid?"
"For example, in the process of designing a new propeller this time, we used a large number of non-isotropic rational B-splines and T-splines. In this scenario, the quadrilateral grid has a tensor product structure suitable for spline fitting. It is easier to obtain tensor product patches than to obtain Bostain/Bezier bases based on triangular meshes, so it should be more suitable for modeling higher-order surfaces.”
Immediately afterwards, Xu Yang placed a notebook with a page of calculations on it in front of Chang Haonan and continued to add:
"After she told me about this idea, I did a simple calculation and found that it should be possible to accurately process the quadrilateral mesh through a multi-point flux finite element method based on the lowest-order Brezzi-Douglas-Marini space. The following work I want to establish the theoretical analysis of velocity superconvergence of this method on h^2-consistent quadrilateral grids, but this is not an area I am particularly good at, so I want to see if you have any ideas. "
The calculation process written by Xu Yang is the same as the program Chang Haonan would write when he was just reborn, with one emphasis on conciseness and conciseness.
It might not be possible for another person to understand.
Fortunately, Chang Haonan's current theoretical level is enough to skip some of the less important parts in the middle.
"Forget it, you should take this book first. Anyway, our top priority is to build the software framework according to your current thinking. The problem of the quadrilateral grid should be a matter of the subsequent product iterative upgrade process. It is not very important yet. It's urgent. It's not too late to tell me when you have an idea.” Xu Yang looked at Chang Haonan who was bowing his head in thought. He also knew that problems in the field of computational mathematics were mainly based on sudden inspiration, so he couldn't rush it at all. So I prepared to take Yao Mengna back to her seat.
In fact, even the grid processing method she proposed is completely new to the field of numerical analysis.
Generally speaking, this takes more time.
Why do large papers in the field of mathematics often take several months or even more than a year to be reviewed?
Because the various new concepts and theories proposed in it are often enough for reviewers to understand for a while.
Sometimes all the reviewers can't understand it, but they can't falsify it, so they can only put it aside for the time being.
However, Chang Haonan called them directly from behind:
"Etc., etc…"
“Although the quadrilateral grid matter will indeed have to wait until later, I still have some ideas about your theoretical proof problem.”
“You understand now?”
Xu Yang turned around suddenly, his face full of surprise.
“It's okay, because I have studied the multi-point flux approximation method, so I know a little bit about it, but it's not a zero-level foundation.”
Chang Haonan directly picked up the pen from the side:
“You can refer to the super-convergent idea analysis of hybrid finite elements. First, find the interpolation that infinitely approximates the exact solution in the finite element space, and establish the relationship between the interpolation and the exact solution, and then establish the relationship between the numerical solution and the interpolation..."
At the same time, at the back of the conference room, Song Jian patted Li Jie on the shoulder and motioned for the latter to follow him out.
Under the guidance of several security guards, the two came to a secluded and uninhabited place nearby.
“Li Jie, what do you think of the prospects of Comrade Xiao Chang's project team, including the company he will establish later?”
“Unlimited.”
Li Jie answered almost without thinking:
“Not only is he an epoch-making leader in technology, but more importantly, I think Comrade Chang Haonan has a very strong ability to lead the overall situation.”
Such an evaluation even far exceeded Song Jian's own expectations, so he became interested instantly:
“Oh? How do you say it?”
Following that, Li Jie roughly described the situation when Chang Haonan met with him a few days ago.
"interesting…"
When he heard Chang Haonan expressed his hope to indirectly promote the application and development of domestic high-performance servers and supercomputers through his work, Song Jian also showed an unexpected expression.
“No wonder Lao Ding attaches so much importance to it and refuses to let go...”
The voice of the second half of the sentence was very low, and Li Jie couldn't hear it clearly for a while:
“Uh... Director Song, what did you just say?”
“Oh, nothing.”
Song Jian waved his hand, then changed the subject and asked quite seriously:
“Comrade Li Jie, what do you think of the reform of the Institute of Computing Technology in recent years?”
This somewhat sensitive question made Li Jie not sure how to answer it for a while.
Since about four years ago, the Academy of Sciences has piloted an enterprise-institution joint approach to entrust the computing institute to the management of Dream Company. This is why the general manager of Dream Company can currently serve as the director of the Institute of Computing.
“There are only two of us here, you can say whatever you want without any psychological baggage.”
"this…"
Li Jie hesitated for a moment.
Some things are definitely not easy to say in public.
But now there are only two of us, Song Jian and myself, so it doesn't hurt to reveal more:
“Director Song, to be honest, I'm not very optimistic about it.”
"When we are entrusted to Dream Company, it seems that the relationship between the two is closer, but in fact it has little impact on daily work. The funds for the three major projects in the institute still come from state appropriations. No matter where Dream is, He was not very interested in our technology development or product promotion. Instead, he was very interested in Shuguang Group. He even tested my tone and asked if the vice presidents of both parties could concurrently serve as the deputy general managers of the other company. "
“And you know, because of Academician Ni's incident, Dream Company has been at the forefront recently, so we have actually taken on some unnecessary pressure..."
"Um…"
Song Jian stood in front of a window with his back to Li Jie, so that the latter could not see his expression.
But it was obviously thinking.
After a long time, the former finally looked back:
"After you go back this time, find a few trustworthy comrades who will be responsible for communicating with Xiao Chang. You must record all relevant information clearly and clearly. After a period of time, we will calculate the input-output ratio. Condition."
This order made Li Jie's eyes tremble suddenly.
If nothing else happens, next year, that is, in 1998, the Academy of Sciences will make another large-scale organizational adjustment.
Song Jian obviously has new ideas.
(End of this chapter)
Chapter end
Report
|
Donate
Oh o, this user has not set a donation button.
|