After this sudden episode, Chang Haonan turned over the blackboard next to him and continued to explain along the route just now:

"Because of the problems faced by traditional methods, in our program, we will use a Delaunay-AFT-based parallel constrained tetrahedral mesh generation method to generate a dividing layer around the partition plane, so as to decompose the mesh coverage area."

"When generating the separation layer, the division plane is only used as a guide, so as to avoid the introduction of geometric constraints, and at the same time, the efficiency of parallel mesh generation is ensured by the multi-level region decomposition strategy. Saves computer memory usage through a disk-cache-based communication mechanism......"

"This process is mainly divided into the following steps, one is the evaluation of the grid size......

"The second is the introduction of the segmentation plane, which can be determined with the help of the principal axis of inertia with the smallest moment ∏of inertia of the center of gravity of M. The direction vector n of this axis can be determined by the eigenvector corresponding to the minimum eigenvalue of the inertia matrix of M, which is calculated as follows......"

“……”

Just the "most basic" parallel surface triangle/quadrilateral mesh generation method in Chang Haonan's mouth took nearly two hours to talk about.

And that's just the theoretical part.

"During the period of preparing for the laboratory, I briefly verified the mesh generation method I just talked about, and the test objects are the common gear hob model and cylinder head model, with 348 and 2736 surfaces respectively, including a variety of mesh sheet surfaces that are difficult to generate in the traditional way, and the final result I put here for your reference."

Chang Haonan clicked the mouse, and finally released the last picture of the first PPT——

Because he had so much to talk about, that during the previous testing process, the office software, which was still quite early, often froze, and he had to divide it into several files before proceeding.

"In this, I define the element mass Q as the ratio of the radius Ri of an inscribed ball to the radius Rc of the receiving ball, and in order to make the mass of the element as an integer as possible, I multiply the result by three, so the final definition Q=3*(Ri/Rc)"

"It can be seen that compared with the high-precision mesh generated by the traditional method, the mesh generated by the new method does not generate a garbage mesh with a mass of 0.1-0.2 even in the dirty area with an extremely complex shape, and the total proportion of the mesh with a mass of 0.6-1.0 is more than 91.7%, and the weighted average mass is about 0.2% lower than the former, which has almost no impact in practical work."

"In terms of final time consumption, although the parallel strategy adds three steps of region decomposition, mesh encryption and node merging compared with the traditional method, the total time for the two test models is only 750 seconds and 2100 seconds, respectively, which is only 5.9% and 4.7% of the serial method......

When Chang Haonan posted the program running log and told the results, the originally silent conference room exploded almost instantly.

The former also deliberately did not maintain order immediately.

It always takes some time for people to calm their emotions after suddenly receiving information that is far beyond their cognitive level.

After all, that's 20 times more efficient.

In the field of numerical computing, a quantitative change in time is itself a qualitative change.

The gear hob and cylinder head are just two fairly simple models, and it may not make much difference whether the mesh generation takes 20 minutes or 6 hours.

But for some super-large engineering projects, there is a world of difference between half a month and a year......

What's more, some more careful people have been able to see that the new method developed by Chang Haonan is better for the more complex models.

In other words, in practical application, this advantage is likely to be more than 20 times.

Li Jie, who was sitting in the back just now and taking notes calmly, raised his head almost subconsciously after hearing the two data Chang Haonan said.

He looked up slightly, and looked at the curtain hanging on the wall in front of the conference room along the lower edge of the lenses of his glasses.

Judging from the enlarged operation log above, it is not mistaken.

That's 20 times.

"This ......"

Li Jie turned his head to look at Song Jian, director of the Science and Technology Committee, who was sitting on his left.

Then he found that the other party was also looking at him at this time.

As absolute bigwigs in the industry, both of them clearly see the potential of this software.

Although what Chang Haonan is talking about so far is only about 10% of the entire project, it is enough to see the superiority of its final product.

Definitely an epoch-making product!

Since 1983, China's first supercomputer Galaxy 1 was put into use, and the main problem that has plagued the computer field in China has been that the computing power cannot be improved.

It's not even our turn to discuss component performance.

But even if others export the same chips and memory to you, the performance of the finished product you take out is still very different.

So much so that in recent years, the United States has even liberalized export controls on electronic components, but it has been very strict in terms of platform architecture.

In short......

I don't think you're a threat.

The reality is indeed similar.

The Computing Institute is the latest Dawning 1000A supercomputer developed by the Liaohe Oilfield, with a peak operating speed of only about 6 billion times per second, and the Dawning 2000-1, which is still under construction, has a target of only 20 billion times.

And ASCI Red, which was just put into use on the other side of the ocean last month, has reached a speed of 1 trillion times per second.

As for the mass-produced commercial models purchased by major universities, the performance gap is even more obvious.

Such a gap is also the main reason for the delay in the development of domestic commercial applications in the field of computer applications.

After all, a limited number of topics can fill up the computing power of supercomputing, resulting in only the most important national projects being able to allocate resources for use, and most researchers are not blessed at all.

Now, with a supercomputing boom led by Beijing Airlines, although this situation has been alleviated to a certain extent, these mediocre commercial servers often seem powerless in the face of complex problems.

If you want to change people's old attitudes, you have to really make people feel the benefits.

At this time, Li Jie's mind recalled that when he met Chang Haonan a few days ago, the latter also said that the business of supercomputing or server cannot be sold, especially in Huaxia, where the related industry is not yet developed, and it is necessary to provide continuous services and promote supercomputer applications.

At that time, he just thought that this idea was very good and would definitely improve some user stickiness, but the computing institute was currently unable to do this work on his own, so he entrusted it to Chang Haonan.

In exchange for the right of first refusal to use the future Dawning 2000 computer, as well as the opening of investment licenses to the Dawning Group.

But I didn't expect that his side hadn't even had time to convene the board of directors, and the other party had even come up with the results.

While there are thousands of fields that require supercomputers, numerical simulation and simulation are definitely the most widely used of them.

The direct improvement of efficiency by 20 times is equivalent to narrowing the computing power gap between Huaxia and the United States from three orders of magnitude to one order of magnitude.

Although it is not enough to directly overtake, at least you can see the back of the other party.

And, almost all engineering disciplines can benefit from it.

"Comrade Chang Haonan, I have a question I want to ask."

A calm voice cut through the noise in the conference room.

Almost all of them looked back to the end of the room.

(End of chapter)