Good progress has been made in the project. Duke can focus on improving the Titan II virtual system. Kerry realized the Titan II in its original style according to Duke's instructions, but Duke felt that this was not the best choice.

Especially in terms of iOS and storage, Kerry has conditions that are not available to the current technology. If the original restrictions are eliminated, the Titan Ii virtual should have better Xìng power than the prototype. Sure enough, when Duke instructed Kerry to make further optimization and improvement, just by letting go of iOS reading and storage allocation, the Xìng power of the virtual Titan Ii almost doubled!

This shows how much impact io and storage have on computing speed, but this is not finished yet. Considering that CPU and GPU have their own strengths, Duke hopes to dynamically adjust the number of computing nodes of CPU and GPU to achieve the best xìng energy. Once this adjustment idea comes out, the Titan II system optimized by Kerry is even more amazing. In testing applications that focus on floating-point operations, the Titan II energy with the largest number of GPU configurations has almost increased by ten times.

Of course, for such a large number of floating-point computing conditions, the gpu configuration ratio has reached almost 90%. You should know that this is achieved without modifying the system architecture at all. It would be strange if the experts who develop supercomputers knew that there was such a powerful simulation system, it would be strange if they did not crash.

You should know that in common research, it usually takes at least 2-3 years to increase the power of supercomputers. After the relevant hardware subsystems can be improved, these higher-energy subsystems can be realized. It is really too evil to realize them after dynamic adjustments like Kerry.

Of course, if you are logically judging more artificial intelligence applications, Duke found that the ratio of CPU to GPU must be reversed to achieve better xìng power. Because this is CPU's strength in this regard.

After this test, Kerry mastered the methods and calculation formulas for dynamically adjusting the virtual Titan II system according to the application type, and he could quickly dynamically adjust the system composition according to the calculation needs of the software system. With this alone, Duke can get the results more than ten times faster than others under the same conditions.

Moreover, Kerry can further increase computing nodes and continue to strengthen the computing power of the virtual Titan ii. Although Duke did not instruct Kerry to make similar improvements, mainly because there is no such demand yet. After Kerry improved the dynamic changes in the hardware, he began to target the cracklinux system for optimization.

On the one hand, this is to improve the system's capabilities, and on the other hand, it is to accumulate experience in system research and development. The major special research project for system research led by President Liu of the University of Science and Technology has now been approved for approval. The first phase of the state allocates RMB 60 million to conduct experimental research on system research.

As a major architect, Duke needs to propose an architectural concept for the new operating system. In the project requirements, this future operating system needs to have four high-energy, high reliability, high security, and high scalability. In addition, it must have the ability to run under a general platform, that is, it must be compatible with current mainstream hardware, including the current mainstream CPU/gpu/memory/hard disk/peripherals, etc.

With such high requirements, the best model that Duke can choose is Linux. However, although Linux is an open source system, because of this, any new improvements based on Linux require to be announced in the world. If you want to publicly promote commercial use, this is unacceptable to the country.

Therefore, there is no problem with the new cào system using Linux as its blueprint, but it cannot be copied in the architecture, otherwise it will infringe its intellectual property rights.

Independently packaged hardware is indispensable. Duke's newly developed mobile phone system does this. However, the mobile phone system is a closed system for a limited hardware platform, which greatly reduces the consideration of compatibility, so it is very efficient.

However, this new system cannot be done like this. In the future, it will have to target specific hardware platforms. Such a system is called a general-purpose system. Duke is embarrassed to say that it was developed by himself. Therefore, the relationship between independent packaged hardware as the core and the new device driver home is difficult to deal with.

Opening up third-party reading and writing of core parts will cause serious security risks. Many viruses will simulate new device drivers to perform core-level operations on the Cào operation system. So this problem is basically difficult to choose now. www.piaotian.comg has to open a gap. Duke didn't know how to avoid this problem.

It was not until Kerry virtual success that Duke was inspired: he looked at the entire basic hardware at a fully encapsulated kernel, and the peripherals must develop drivers by calling the kernel's SDK to realize data exchange and operation calls between peripherals and kernels. In this way, Duke can ensure the security of the kernel.

Compared with the endless peripherals, the environments that the kernel needs to target is very limited. There are only two or three CPUs. Although the total number seems to be quite large, the total number of models with real changes in the instruction set is no more than 20 instruction sets!

But after all, they are limited. For example, Intel's general desktop-level CPU instruction set. As of 2014, the extended instruction set part was only nine types such as mmx, sse, sse2, sse3, ssse3, sse4.1, sse4.2, sse4.3, em64t, etc.

Therefore, with experience in high-end CPU simulation, it is not difficult for Kerry to simulate relatively low-end CPUs, it is nothing more than reducing xìng power and instruction set support.

Duke can develop a core processing application for a higher-end CPU. The remaining compatibility will be revised on this core application, and the instruction set is compatible with the design based on different CPUs.

With the supercomputer system as the basis, Duke will initially consider distinguishing this system into two models: desktop and server.

The server model will support the massive memory and super CPU clusters of the existing supercomputer architecture, realizing the unity of mainframes and medium-sized workstations. For many applications, such as databases, www.piaotian.comg.

The desktop model will still support multi-logical CPUs, but the number will be controlled at 8-channel physical CPUs. According to the main CPUs in 2014, it is 256 logical CPUs. In this way, even a small workstation with 8-channel can be fully utilized in this system.
Chapter completed!