For no other reason, the pre-research report of "Juxin No. 1" far exceeded everyone's previous expectations.

As the first chip of matrix semiconductors, in most people's imagination, "Rectangular Core No. 1" should be independently designed on the basis of ARM instruction set.

In fact, just a week ago, Matrix Digital spent $10 million to buy hundreds of ARMv7 instructions from ARM.

However, in this report, the instruction set of "Moment Core No. 1" has almost nothing to do with ARM, although there are still some ARM instructions in it, but a large number of instructions with new complex instructions (CISC) and ultra-long instruction sets have been added, and they can see that these instructions are obviously original content, and some concepts they have not touched at all.

In a sense, this is already a new set of instruction set system that has departed from ARM's reduced instruction set structure.

Everyone can't help but look at each other, you must know that Apple and Samsung, although they also launched self-developed chips on the basis of purchasing architecture-level authorization and core-level authorization from ARM, they are still under the ARM system.

First of all, we need to clarify two concepts, what is architecture-level authorization and kernel-level authorization (also known as IP authorization).

Architecture-level authorization refers to the ability to greatly transform the ARM architecture, and even expand or shrink the ARM instruction set, Apple is a good example, Apple uses the ARMv7-A architecture on the basis of expanding its own Apple swift architecture, thus launching the A6, A7 chips.

Kernel-level authorization, on the other hand, means that you can base on a kernel and then add your own peripherals. For example, USART, GPIO, SPI, ADC, etc., and finally formed their own MCU.

Samsung TI has obtained the architecture license and kernel-level core license from ARM. This means that it can create its own kernel architecture based on the ARM instruction set as needed, and can add various on-chip and external components such as communication interfaces, display control interfaces, GPIOs, and so on. Thus producing its own "processor chips".

It's like if I write an article and I tell A that you can use it after you modify it, that's architecture-level licensing, and I tell B that you can quote my article in your article, which is kernel-level licensing, and I tell C that you can only forward my article. It can't be changed, and it can't be added, so it's a hierarchical license.

And the matrix semiconductor, not even a reference, it only borrows some concepts from the ARM article, and the central idea has nothing to do with the ARM article.

Xia Ye, the deputy general manager of Matrix Semiconductor, glanced at Yang Lin in surprise, half a month ago, when Yang Lin asked him to go to the United Kingdom to buy an ARM license, he also wondered why Yang Lin only let him buy a license for hundreds of ARM instructions, and also offered a sky-high price of 10 million US dollars, you must know that this can buy a complete chip design scheme from ARM.

But now. He vaguely understood.

The reason why Yang Lin spent a huge amount of money to buy the license of ARM is probably to ensure that MatrixOS can run on Android phones in the future, and on the other hand, to avoid disputes over ARM copyright in the future.

But why did Yang Lin do this?

Summer Night frowned.

The uproar caused by the mixed instruction set lasted only a short time, and the crowd quickly continued to watch.

They are slow to watch.

Although this report is only a dozen pages. But it is full of dry goods, and every sentence and concept involves countless basic theoretical content. Most people look at the R&D personnel who have a weak foundation in software. I didn't even understand the specific meaning of some of the instructions.

But one thing they know very well is that the level of this report is very high. There are many brain-opening concepts mentioned in it that they have never heard of, but when you think about it, it feels logical.

They haven't experienced this feeling for many years, and after working in the field of chip design for so many years, many people have become screws in engineering design, and they can complete their tasks under the arrangement of the team, and they don't need too many other ideas at all.

And this report seems to open a window for them, so that they are fortunate to be able to witness the sky outside the window.

This feeling is even stronger when it comes to the design part of the compiler.

The importance of the compiler to a chip is undeniable.

The programmer writes the programming language when programming, but the computer runs in the machine language, and the compiler is a tool that translates the programmer's programming language into the machine language, and the quality of the compiler also greatly affects the performance of a computer, for example, SUN once relied on the optimization of the compiler to increase the spec score of the CPU by 50%, which greatly improved the performance and user experience of the whole machine.

At present, the most widely used compiler in the world is GCC, which has different CPU optimizations for X86, ARM, MIPS, Alpha and other instruction sets, and is better optimized for X86 and ARM, which have a high market share, and the version is relatively new (part of the code of GCC is provided by Intel), but the optimization of Loongson and Shenwei is relatively poor, and the version is relatively old.

But the new compiler mentioned in this report is tailor-made for the previous hybrid instruction set, and there are many very interesting ideas and concepts that these IC engineers are fascinated by.

Many people understand why the company signed a confidentiality agreement with them before, and the content of this report alone is enough to cause a huge earthquake in the IT industry around the world. Not to mention that they have to design a new product based on it.

Before you know it, two hours have passed, and all that remains in the conference room is the clatter of pages turning and discussion.

Yang Lin was not in a hurry, only occasionally asked Bai Hua to add some water to everyone, and spent most of the rest of the time chatting with Qiao Zhenyu in a low voice.

It wasn't until three hours later, seeing that everyone didn't see much more, that Yang Lin smiled and said, "Okay, presumably everyone has almost consumed this report, do you have any special ideas?" ”

Everyone, you look at me, I look at you, and no one speaks.

After a while, Xia Ye raised his hand and said: "Mr. Yang, I have a question, what is this Rectangular Core No. 1 we want to design for, I'm afraid ordinary operating systems can't use such a design, right?" ”

Everyone's eyes were focused on Yang Lin, which was not only the question of the summer night, but also the question of everyone at the scene. (To be continued......)

PS: Ask for a subscription, ask for a monthly pass!