Shulou(Shulou.com)06/01 Report--

What this article shares with you is about the utilization rate of CPU in the single-chip microcomputer. The editor thinks it is very practical, so I share it with you to learn. I hope you can get something after reading this article.

Open the computer's task manager, look at the beating CPU utilization, and find it very comfortable. Each thread takes up a clear number of CPU, so you can really think that your computer is running slow.

Today's note does not talk about the use of each task (or thread) CPU, but the overall use of CPU of single-chip microcomputer, easy before difficult.

Why do you need to know this? What's the use of knowing this? Osprey reads few books, so it doesn't write official words, but only talks about its own understanding.

The higher the CPU usage, the busier the system, and the slower it responds to some things. For example, your computer CPU usage accounts for more than 90%, you will find that typing is slower, mouse movement is slower, this is all because the CPU is too high, so that the system does not have time to process your keyboard and mouse data, so there will be slow performance.

The computer is a non-real-time system, the requirement is not high, even if the computer slows down and the computer crashes, the consequences are not very serious, but if your embedded system is in the field of national defense and medical treatment, if this happens, the consequences will be unimaginable. For example, if there is a sudden problem with the ventilator, it is a disaster for patients, so the products of the medical industry will undergo rigorous testing, otherwise they will not be allowed to go on the market.

Most of the embedded systems should be real-time operating systems, that is, the so-called RTOS, which must respond very quickly to external situations. If not, there is something wrong with the system you design.

So how to respond to external information quickly? Depending on the usage of CPU, the lower the utilization rate of CPU is, the faster it can respond. How do you understand this sentence?

For example, in a day, you have to work for 8 hours, and the rest of the time belongs to you. If calculated on a daily basis, your CPU utilization rate is 80.2433.3%. You can quickly respond to other things at other times, such as when someone asks you to go out to dinner. If you are on call during off-duty hours, it is useless to call you. You can only wait until after work. So although your CPU utilization is only 33.3%, you still can't respond to other things in time at work, because going to work is the highest priority task (assuming that going to work is the highest priority task).

This example may not be very good, but it may be better to replace it with the example of students. For example, a student has 7 classes every day and has a break between classes, assuming that he or she will take 8 hours, but because the middle is not continuous, although your CPU utilization rate is still 33.3%, you can always respond quickly to other things during recess, so the overall performance may be better than the previous example.

So when designing the system, do not let a high-priority task occupy CPU for too long, if possible, split the long task as much as possible, otherwise the low-priority task may not be able to run in time, the external manifestation is that the system is stuck.

After watching this, a lot of people will think, how should my system calculate CPU utilization? By the way, my system is bare metal.

I'm sorry, the utilization rate of bare metal system CPU is 100%, which doesn't count.

Well, what about those with an operating system, such as uCOS, FreeRTOS, RT-Thread?

Strictly speaking, if mechanisms such as hibernation are not used, CPU usage is 100% in both bare metal and operating systems.

Why would you say that? You can see the method of calculating the CPU utilization of the system. (here we are talking about the simple calculation method in RTOS, not the computer, which should be more complex and Osprey is not clear.)

To put it simply, there are many user tasks in an operating system, and there is a special system task, that is, idle tasks. This task usually does not dare anything, just run there, CPU will run to the idle task when there is no other task to perform.

Besides running, what are the characteristics of idle tasks? With the lowest priority, it is not allowed to suspend an idle task, that is, the task is always ready.

Because of these characteristics, it becomes very special and is at the core of our ability to calculate CPU usage.

To put it bluntly, the so-called CPU usage calculation is to calculate the elapsed time of idle tasks first, and then deduce the elapsed time of other tasks.

For example, if the idle task runs for 700 milliseconds in 1 second, the CPU utilization of the idle task is 70%. If you push it back, the utilization rate of other tasks is 30%. But pay special attention to the fact that idle tasks run for 700 milliseconds, not that idle tasks run continuously for 700 milliseconds, but that they are interspersed with the execution of other tasks, which is interspersed with 300 milliseconds to execute other tasks.

Just look at this picture and make it clear:

In fact, there are far more task switches in 1 second than shown in the image above, and it is only for better illustration that there is not so much switching process.

In a really good system, a task will not occupy CPU for a long time, but will always take the initiative to hand over the right of use. For example, Task 2 above takes up 100ms. If this is a high priority task, then the response of a low priority task must be more than 100ms.

Of course, if the response time meets the design requirements, it doesn't matter when the number of tasks in the system is relatively small.

Maybe you still have doubts, why don't you talk about idle tasks, which have 300ms of CPU usage?

Sorry, I really don't need to talk about it, because it has the lowest priority task, so if it can run continuously for 300 milliseconds, it must be because there are no other tasks to deal with to keep idle tasks running.

Why would you say that? Because in the operating system, in addition to active switching tasks, there is also passive switching.

The so-called active switching task means that the task itself thinks that it has finished its execution, and then actively calls the system function to switch, such as the system delay function, etc.; while passive switching is different, passive switching occurs all the time. As long as the conditions are met, then your task may not be fully completed, and you may switch to other tasks to be executed first.

How do you understand it? If four people form a group to discuss the problem, one of them is the group leader (operating system). The group leader has the absolute right to interrupt other members (tasks) at any time, so when the team member speaks, he will check every few minutes to see who raises his hand to speak. Once a high-level member is found to have raised his hand, no matter whether the current speaking member has finished speaking or not. The group leader will immediately let the high-level team members speak first, and only after he has finished speaking will the group members who have not finished speaking before continue to speak. This may not be very human, but it does guarantee high efficiency!

In fact, there is one very important thing that has not been drawn in the above picture, and that is the checking of ready tasks by the operating system from time to time. In the operating system, this kind of checking is usually done by timing interrupts (stm32 has timing interrupts specially prepared for the operating system, that is, SysTick).

Interrupts are super tasks that take precedence over all tasks (or threads).

However, the inspection time (that is, interrupt time) is also fastidious. If the inspection time is too short, the whole system will be busy switching tasks, and the proportion of time spent on task switching will be very large; while the inspection time is too long, then high-priority tasks will not be able to get a faster response, so this time must be chosen carefully.

Generally speaking, task switching CPU should be less than 1% (this has no theoretical basis, ha, written by Osprey), that is, if each task only calls one delay function, if your CPU is in this range, then it is more appropriate.

When you learn CPU usage calculation, try to modify the interrupt time, you will find that different interrupt time, CPU usage is different, the reason is the consumption of the operating system itself!

The above is what the CPU utilization rate is inside the single-chip microcomputer. The editor believes that there are some knowledge points that we may see or use in our daily work. I hope you can learn more from this article. For more details, please follow the industry information channel.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.