As a chip engineer, from the first day of graduation to the first day of entering the company, it is necessary to complete an obvious transformation. With the development of the work and the progress of the project, the personal ability has also been subtly improved. When we look back When looking at the growth curve of personal knowledge/skills, you may find many occurrences and changes, which are not the same as your original plan. Looking forward, you will also find that the future requirements and the skills in our hands seem to be still very big. difference, so what kind of learning curve is reasonable and scientific? I would like to share/discuss some of the author's simple opinions here, and please correct me if I am inappropriate. Without further ado, ICer GO!
desire to grow
Everyone is eager to grow up and grow up, not just height/appearance, but also skills and knowledge, self-realization continuous breakthrough, of course, this is also your source of happiness (one of the sources of endorphins) (from: How to get More endorphins? )
Just like when you were in school, you solved a difficult problem, completed the additional questions of the exam, and got grades, likes and good rankings from your classmates; you also need to do similar things at work, To show your value, as a chip engineer, it may be: using your own algorithm to realize the writing of a convolutional verilog; or using UVM to complete the verification of the interconnection of the CPU claster; or a TCL script to complete the manual DFT chain Fine-tuning and so on, all in all, get rid of the appearance, students are still "doing problems".
Learn the basic knowledge, master the basic theory, and then use your own code to complete small tasks one after another on a basic platform by learning the existing processes, scripts, and documents until the project is successfully TO out. When the film is returned, your work results are recognized by the chip test, and a hanging heart falls, Yes, I Can! . Congratulations, you have reached another level on the spiral staircase.
Correspondingly, material changes such as appreciation and salary increase have brought you not only endorphin happiness, but also hard physical objects that can announce good news to your family/girlfriend.
This is simply the growth curve through the project, which has the following elements:
- project based
- Collaboration
- Individuals focus on their own business/work
- Overall success can only reflect personal success.
The following things are even more similar, constantly spiraling upwards, forever.
Personal value is also gradually enlarged, and the sense of self-identity to oneself is also gradually strengthened. Understand the technology, do the work well, don't lose the chain, and slowly integrate into your blood, and then a standard technician will be born at this time!
platform and individual
Chip design is a complex engineering project with a relatively long business line. Basically, everyone is very busy because of the deadline. But very few projects start from scratch. Even for new projects, many IP resources and EDA processes will be provided. Of course, people are still inseparable in the end, because talents are the core competitiveness of the company. Here we need to separate out the factors that make the project successful: the platform that favors hardware, and the manpower that favors softness.
platform value
Simple things rely on creativity, and complex work depends on processes. Chip design clearly falls into the latter category. Whether it is front-end/middle-end/back-end/DFT, etc., each stage has a set of independent processes. What is a process: A process is a specification to avoid personal errors and accelerate products . If it is said that the company set up stage engineers to sing operas, then this process can be simply regarded as this stage . The iron-clad battalion is full of flowing soldiers. This chip design is part of this process in the battalion.
The bigger the company, the more attention it pays to the process, and the basic process is also the golden book of silicon proven. This point can be felt by ZTE if you look at foreign manufacturers or HiSilicon. If the backend is considered, the process can also be distinguished according to different fabs/nodes.
Because the chip design is too complicated and the process is not built from scratch, the current popular method is to build it based on EDA tools.
| design steps | Process points | auxiliary tool |
|---|---|---|
| code design | code specification check | linting/spyglass |
| Functional Verification | Simulation environment | vcs/nc-sim |
| DFT | Testability design and implementation | text/text |
| Middle and back-end implementation | Comprehensive/PR etc | DC/ICC/PT/Calibre |
All of the above are standard tools provided by EDA companies. If users want to use these tools, they need to build a process in each tool. The approximate example is as follows:
If one of the steps is expanded, it will probably be a diagram like this
But here comes the question, where area you?
OK, found it, congratulations, you are doing APR! If you look at the whole picture, the engineers are not really working for the company, they are working for the process
Therefore, strictly speaking, the technical composition of the project should be: engineers use the process to complete project tasks.
But note that because every project has to go through the process, the power of the process will become more complex and powerful, the automation will be higher, and the convergence will be stronger. If there is a simple formula as follows:
项目技术总需求 = 流程 + 本项目的工程师的代码量
Then, every time the project is completed, the engineer's code amount will be packaged (archive) into the process script, so that larger and more stable projects can be done in the future, so it can be seen that the development amount of an engineer's nth project is :
项目n技术总需求 = 项目n-1的流程 + 项目n的工程师的代码量
As the design becomes more and more complex, the importance of the process is self-evident. This is where the major EDA companies are currently attacking. What's more, they have brought AI to play with them recently, so it can be seen that the current and future trends It is inevitable that the amount of code of the engineer will be swallowed by the process (doing a tug-of-war with the engineers of the EDA company?). From the
perspective of project development, the purpose of doing this is very reasonable: the success rate of the project is the first priority.
If this model is compared according to the company's honesty, then the scale of the same project is roughly like this:
It seems that the total amount of code is similar, so why start-up companies are tough on established companies, the answer is no, this model can be refined as follows:
项目n技术总需求 = 项目n-1的流程*准确率 + 项目n的工程师的代码量*准确率
Since the quality/completeness of the process is much higher than that of manual labor, in the end, it is difficult for the same engineer to produce projects of the same quality in the new company's process. This is not a human problem, but a platform construction problem.
In projects in a complex working environment, the value of the process is far greater than the value of people. This is only from the perspective of EDA automation. If you look at it from the perspective of IP, PHY, coding, etc., this phenomenon will be magnified even more.
Therefore, the platform may not depend on most engineers, but there is no platform, but most engineers cannot do without the platform
personal value
The code of the project still needs engineers to code. Without engineers, the platform can produce nothing. If the platform is a river, then the engineer is a boat in the river. Without a boat, nothing can be transported. But note that the ship is dynamic, and the choice of engineers is also free.
Due to the high platform, large space, and strong capital of a large factory, usually the peak is when you enter the job, and you can have a feeling of seeing all the mountains and small ones.
It's still an old saying: Dasi is fragrant! Due to the difference in platform construction, large companies and small companies have completely different requirements for people
| company | process | Own work | job fill |
|---|---|---|---|
| Daishi | Professional maintenance | focus focus focus | One radish and one pit, whoever does the work |
| Koshi | develop by doing | too many things to focus on | I have to do everything, I can't do anything well |
Dashi climbs the ladder
Dasi’s platform is so good, what engineers may need to do in the past is to keep digging and digging in a segmented field, and iterate and iterate non-stop. Note that you don’t start from scratch, you are based on the foundation of many previous versions , the former planted trees and the descendants enjoyed the shade. Of course, large platforms are also very busy. For example, HiSilicon-style companies have a good platform, but the projects are also huge. Very high. Similar to climbing a ladder:
For large projects, this ladder may be very long, very long, but because of the good platform, you don't have to worry, it will definitely be able to climb up, this is the platform strength of a big factory. For engineers, here are a few features:
- Need to have the basic knowledge of the position, learn by doing
- Refer to/inherit the work of predecessors, there is a high probability that there is no risk of not being able to complete
- If you have any questions, you can ask EDA or colleagues around you, and you can consult with all the resources
- The ladder is relatively long, you must be able to endure loneliness and continuous overtime
Xiao Si builds a house
In a small company, the team is usually not very big, and the project will be relatively small, but due to the imperfect process (usually there is no EDA company support at all), so doing a project is building a house, but note that the leader will not Let you build skyscrapers, because maybe Xiaosi can't even support the cement costs of skyscrapers.
For the first project, let’s build a bungalow first, and see how the team works:
For engineers, here are a few characteristics:
- The team must have someone who understands (that is, someone who has done a complete project)
- There are a lot of chores in building a house. You have to do everything. You have to be able to do everything. If you can’t do it, at least you have to learn it.
- The process is built on site and changed while doing it
- Requires strong initiative, cannot wait and let
- Almost no one to ask, everyone is busy or just needs you to get it done
- There are few iterations in the project, so no one knows whether there will be problems when the house is built. Don’t complain if there are problems. Be patient and caring
The ladder of the big company is long, and the house of the small company is wide. If you look at the area and workload, the difference is basically the same:
due to the difference in platform, the risks of the two are different.
Such differences have obvious differences in the requirements of people. For newcomers or junior engineers, they should choose to stay in Dasi before watching other people build houses, so that the ladder will not fall and climb well. For experienced engineers, both are fine. If the house can be built well, it’s okay to start a small high-level company (the logic of a start-up company). However, if a small company wants to succeed, it must need a big card position , otherwise failure is highly probable.
learning curve
Regardless of whether the engineer is in a small company or a large company, the general direction/process of the entire project is basically similar. Maybe the products of the small company are relatively simple and will be simplified. Returning to the personal learning curve, the subject has some step-by-step learning suggestions as follows:
- Skills Overview
It is important for an engineer to have a technical overview of the position.
The chip business process is relatively long, and any node needs a relatively long process. Engineers are new to the industry. By looking at some systematic data, company processes, tool UG, etc., they must understand the value and characteristics of this process. This is very important for the flow control of future work, and can also provide services for the project more efficiently.
Here is a simple example to illustrate, for example, post-simulation, which usually occurs in the database delivered after the completion of APR. The purpose is to bridge the gap between static timing analysis and gatesim. Generally speaking, it can only be done after APR actually converges. Start gate-sim, if the engineer can accurately distinguish the focus of gatesim and APR ECO, it is completely possible to provide a gate-sim simulation environment in a faster way, which can speed up gates-sim for at least one month.
If you do not even have access to the technical overview of your position in this position, it can basically be regarded as a situation where the company's processes are not perfect or there is a shortage of backbone manpower. There is a high probability that the things produced in this way cannot be guaranteed. - The high probability of technical expertise
is the field of personal focus. For example: RTL design engineer, who focuses on developing and calling PCIE, here needs to spend a lot of effort to read the tome of PCIE several times. Good for digestion and absorption. If you are a verification person, in addition to understanding the PCIE specification, you also need to look at the places in the verification environment/process that affect you. In addition to the bugs in the RTL code, you also need to be careful about the pits of the process. This technical point has been cultivated for a long time, and it will make you an expert in this field in the future. After all, PCIE6.0 is coming, is there anything better than figuring it out? - The understanding of Involution
In English, there are some words with “volution” as the root word, which roughly mean the following meaning “whirlpool”. If it is used as the root word, the following words are common:
| word | paraphrase | postscript |
|---|---|---|
| evolution | swirl | root |
| con-volution | incomprehensible things | – |
| devolution | Decentralization | – |
| e-volution | evolution | – |
| re-volution | revolution | – |
| in-volution | degeneration/involution | and evoluton are antonyms for each other |
It can be seen that the introversion that is often said at present actually has a degenerate meaning in it. Therefore, innovation is not a competition (competition) or a challenge (challenge). It is more of an internal degeneration or a team internal consumption.
The understanding of involution: If a thing (volution) cannot evolve outwards (e-volution), and it is iterated repeatedly by the same people in a similar way only on existing resources, the final thing (volution) will fall behind. into the involution (in-volution).
The core problem of involution is that it focuses too much on the current problem, and places great hope on using existing processes and methods to solve the current problem. The stronger this force is, the more serious the involution will be. One way to get out of involution is to increase the dimensionality. That is, after learning the current method, jump out of the existing method process and optimize from a higher dimension. This requires a strong sense of evolution, even revolution. This is a challenge to oneself as well as a challenge to the team.
Don't fall into the vortex of involution, and don't be proud of involution. All the resources consumed by involution will not become income due to consumption (cost) revenue), because the whole system is in a state of involution.
4. Flat learning
Whether you are climbing a ladder or building a house, this is very important. As mentioned earlier, the platform process is air and water in daily work. Everyone has not felt its value at all, so many people come out of the company. Most of my classmates, especially those from big factories, have very narrow business, so it will be more difficult to work in a new environment or a small company, especially for engineers with long experience. If there is only one ladder in their life, it is difficult to move out. Therefore, it is necessary to establish a flat learning path as soon as possible, especially for the employees of the big factory, otherwise the pressure will be very high if they leave the big factory.
The code solves specific problems, and the process solves project problems. If you want to improve the dimension, you need to look at the problem from the perspective of the process platform
Therefore, while focusing on your own business, you need to pay attention to the process of realizing this business. Indirectly, you can also refer to the processes of other business departments.
Understanding a process is the first step in optimizing a process, optimizing a process is the first step in creating a process, and creating a process is the first step in breaking a conventional process.
Only by flattening learning and improving the dimension of thinking can we jump out of the current vortex (volution) and avoid involution (involution).
Through flat learning, whether it is climbing a ladder in a large company or building a house in a small company, the final knowledge topology must be realized as a large rectangle
If we take into account the actual situation of personal experience: it is impossible for people to be comprehensive and proficient in facts, and the total amount of real knowledge area is similar to a normal distribution
Through flat learning, expanding knowledge, using the spirit of evolution and even revolution, and absorbing resources and strength from the outside world, we can effectively avoid mutual involution in an effective space, so as to achieve personal improvement and team value.