Reborn in 2010, the AI ​​in my mind is too powerful.

On Friday, Chen Yiran finalized three patent applications at Fenwick & West.

Application 1: Application method of incremental computing engine in EDA placement, which is the core algorithm of DeepFlow's dirty mark propagation and topology sorting-driven dependency management to adapt to two-dimensional spatial layout.

Application 2: A spatial index-based placement engine adaptation layer, which replaces the AST tree-shaped triggering conditions with a spatial + connectivity hybrid index, and designs a macro cell and standard cell constraint resolution layer.

Application 3: A reinforcement learning-driven chip placement optimization method, which utilizes a framework of multiple sets of parallel training to simultaneously explore different placement strategies and share experiences on the placement task.

The three patents are essentially locked in a fierce battle, with each link blocking the industry's inevitable path.

This marks their second collaboration, and this time Chen Yiran also signed an agreement on behalf of FlowEDA with the law firm. Fenwick & West will serve as FlowEDA's external legal counsel, providing various legal services.

After returning home, Chen Yiran uploaded all of DeepFlow's core code to GitHub. He also wrote a description: DeepFlow is a high-performance deep learning framework whose core feature is its incremental computation engine, which automatically tracks changes to the computation graph and recalculates only the affected nodes. It is suitable for scientific computing, EDA, and reinforcement learning scenarios that require rapid iteration.

He even sent an email to TOM to say that he had added some great stuff to his website, which was a way of showing off.

Then he started packing his luggage, since he was leaving for Anaheim the next day.

Anaheim is a small town in Orange County, Los Angeles, and is home to the world’s first Disneyland.

Known for its theme parks, sporting events, and convention economy, Chen Yiran wasn't there for leisure; she was attending a convention.

The 47th DAC, or Design Automation Conference.

In the EDA industry, this is a must-attend exhibition.

The DAC started in 1964, more than 20 years before the term EDA was even used.

By 2010, it had become the 47th edition. Every year, it is held for a week in June, and thousands of people participate each year. 40% are from the three major EDA companies, 30% are sent by chip design companies, 10% are startups trying their luck, and the rest are people from academia.

At this time of year, manufacturers of all sizes crowd into one exhibition hall, competing with each other. If you don't show up for a year, your competitors will wonder if something is wrong with your company.

The 2010 edition was originally scheduled for June 13-18, but due to a one-week delay in venue scheduling, it was changed to June 20-25.

Since the company has not yet completed commercialization, it does not need to rent a booth to showcase its achievements and attract investment.

Chen Yiran and Gao Dewei bought the Full Conference Pass, which cost $800 per person. They received a name tag and a free coffee.

On Sunday, the first day of the exhibition, the two went straight to the main exhibition hall.

Synopsys' booth was in the best central location. It had two floors; downstairs, six demo stations were running real-time placement rendering simultaneously, while upstairs were numerous small meeting rooms for discussing collaborations. Representatives from Qualcomm, Broadcom, and Huawei were also present near their booths. If you became interested, you could go upstairs to discuss initial intentions.

Diagonally opposite is Cadence, whose booth is equally impressive.

Every year at DAC, the two companies compete to see whose booth is bigger. It's said that they even compete on coffee machines. One year, Cadence suddenly provided Lavazza coffee machines, and Synopsys quickly went out and purchased several illy machines to display.

The Starbucks at the Mingming Convention and Exhibition Center is free.

On the other side, Mentor Graphics was more low-key. There was a queue at the entrance, with a group of engineers squatting down to help people troubleshoot on the spot.

Seventy percent of attendees went to the Mentor booth to fix bugs. Their Calibre DRC verification is almost a monopoly in the industry. Everyone who makes chips has to use it, and everyone has been screwed over by it. Mentor's booth turns into a high-end customer service center every year.

The three main booths together occupy one-third of the main exhibition hall, while the remaining two-thirds are occupied by small and medium-sized manufacturers and academic exhibition areas.

Small companies' booths are all very similar: a few roll-up banners and posters, a few tables, a few computers, and a bowl of candy on the table.

Among them are many very capable companies, such as AtopTech, Rio Design Automation, and Pyxis, all of which have achieved some success in their respective fields.

Chen Yiran felt that his company would likely set up a booth here next year, with the goal of eliminating all the companies mentioned above.

Knowing yourself and your enemy is the key to victory in every battle. Chen Yiran and Gao Dewei spent a long time sneaking around several booths, and even took a lot of brochures with them before leaving.

At noon, the two went to the academic lecture hall on the second floor.

The DAC technical program has hundreds of peer-reviewed papers, which are scattered in dozens of small conference rooms in the afternoon according to research direction.

Gao Dewei went to the physical design session, where his former colleague presented a new algorithm for multi-level layout netlist partitioning.

Chen Yiran then went to see the Emerging Technologies report on his own.

The conference room wasn't large; it probably had about thirty or forty people seated inside.

This was actually Chen Yiran's first time experiencing cutting-edge technical reports in person, but he wasn't particularly impressed.

After observing and learning around with Jasmine for a long time, we've at least confirmed that DeepFlow's incremental computing engine is incredibly powerful at this stage.

Once it is released, it will definitely cause a major upheaval in this research direction.

Of course, I also learned a lot of new things, not just technical ones, but also more practical industry insights.

Because the reports for the design field are a bit odd, the people from the design companies aren't there to present papers at all; they're practically there to complain and whine.

When a Qualcomm vice president was talking about their MSM series baseband chips, he didn't talk about the technology; instead, he focused on the nightmare of back-end design, loudly complaining about how unusable the EDA company's tools were.

An AMD employee explained their GPU physical verification process: the RV870 has over two billion transistors, and running a complete DRC takes fifty hours. The design verification window before chip fabrication is only five days. If the EDA tools malfunction, half a day can be wasted.

The audience was mostly product managers from various EDA companies. They come every year to listen to complaints and see what their clients are dissatisfied with so they can go back and revise the iterations.

These experiences are still very useful. In the future, when Chen Yiran makes products, she will definitely try her best to avoid making the client feel awkward using them.

It wasn't until the afternoon that Chen Yiran and Gao Dewei met at the coffee corner, chatting about their experiences while drinking coffee.

As a veteran, Gao Dewei said with a hint of pride, "Isn't DAC amazing?"

"That's quite interesting."

"There's a saying that the DAC exhibition hall is the Darwin pool of the EDA industry. When you walk in, you think you're seeing the Big Three and two hundred small companies, but what you're actually seeing is a living food chain."

"Not bad. I've looked at a few mid-sized companies today, and I think we can take them over next year."

"Sure, Synopsys and Cadence are at the top, competing with each other. Mentor is on the side, surviving by monopolizing a niche market. In the middle are companies with proprietary technologies. If we can take some of those in the middle, we can live comfortably."

"You have such a lack of confidence; just surviving isn't enough."

"EDA startups generally don't survive for more than five years. After that, they'll either be acquired or go out of business. Just surviving is already quite an achievement. But our company's ideas are indeed more innovative than theirs. I feel that we won't go under in five years, but we'll most likely be acquired by a giant."

Chen Yiran didn't respond. In fact, from a normal business perspective, operating such a company long-term would be far too uneconomical for a reborn individual.

Jasmine has a complete memory of his life in her mind, and there are many ways to make money.

Putting aside other things, in early 2016, the Powerball lottery jackpot reached a peak of $16 billion. Chen Yiran was in Los Angeles at the time and joined in the fun, queuing up to buy lottery tickets a few times. He remembered all the winning numbers from the last few draws.

In the end, on January 13, 2016, three people split the jackpot. One of the winners was in Los Angeles, only one kilometer away from where Chen Yiran lived at the time.

However, Chen Yiran is not obsessed with money.

It's not that he doesn't care at all, but his core goal is still to develop AI and brain-computer interfaces, so he definitely can't just sell the company like that.

Taking this opportunity, Chen Yiran also revealed his lofty ambitions to Gao Dewei: "Although it may sound a bit boastful now, my goal is to make better chips and more powerful AI. Therefore, the company will absolutely not be sold. You don't need to worry about the company's survival. What I want first and foremost is the continuous iteration and development of technology. Survival has never been a problem for me."

Gao Dewei nodded, not entirely believing it on the spot. He was over fifty and had heard far too many empty promises, so he was used to them.

"I suggest you talk more about your lofty ideals at the company; it's very inspiring. After all, I'm getting old and don't have that much drive anymore."

"You own shares in the company. If the company really succeeds in the future, your grandson will never have to worry about anything for the rest of his life."

"That makes sense, and I instantly feel motivated again."

Chen Yiran was sipping her coffee, about to continue dangling promises to Gao Dewei, when she caught a glimpse of a booth nearby. Because of its secluded location next to the coffee corner, she hadn't noticed it while casually browsing earlier, but now she saw it.

That was SMIC's booth. Chen Yiran threw away his paper cup and prepared to go over and see this future partner.