Some thoughts on game performance optimization

Recently, when doing optimization, I feel more and more , Optimization is not to add a comprehensive gain to the game Buff, The essence of optimization is to make trade-offs , On the one hand, it optimizes , It will inevitably bring about another problem , It is a zero sum game .

For example, we often use cache or object pool to improve the performance of the game , Reduce Caton , This is a typical space for time . At the same time, we also occasionally use compression technology , Compress a file , Or data tables , such as AssetBundle It's compressed , Then decompress it when reading . This is a typical time for space .

We are always changing time and space , Is there any optimization that improves performance from space to time ？ Not without . But the cost of optimization may be engineering complexity , Smarter code often means less direct , More difficult to read , More difficult to expand and maintain , It may also be optimization for specific situations , Reduce the versatility of the solution . At first glance , This optimization seems acceptable . Then I have to give a few examples , Let's have a good discussion about this problem .

For example, a simple sort , We started with bubble sort , Later, it was changed to quick sort , We assume that fast sorting is brainless and better than bubble sorting , Time complexity and space complexity are lower . But at the same time, we also made a hypothesis , That is, the code that is quickly arranged is more difficult to write , It is also more difficult to understand , The students who took over didn't understand how the algorithm was implemented , Dare not change , Then there is a need to modify some logic of sorting , The classmate panicked at once , Finally, it was changed for a week, and finally it was changed , It's hard on the head . If bubble sorting is used , I believe this student will soon be able to meet the needs of new changes .

Here is just one example , But it is enough to explain some problems .《 The code of 》 In the said

The first technical mission of programmers is to control complexity

Everyone has a different understanding of this sentence , What complexity is controlled , But in my understanding , Part of it is code complexity , Architecture complexity , Engineering complexity . Maybe you used a very clever and advanced architecture and Algorithm , But you can't completely control yourself , Then you lack control over it , Maybe this version can run without problem , But you can't face the problems well , Future maintenance extensions , And modify the requirements .

I'm not saying that we can't write better algorithms . It's about expressing , Any optimization comes at a price , But sometimes you don't realize the form of cost , Think this is not brought about by optimization .

For example, I once did an optimization in the project , The result of this optimization is , All resources need to run the script again and modify it before they can be used . As a result, the packaging processing time is too long , It's too late to get out of the bag , It affects the acceptance of the test part , Lead to frequent complaints , Eventually my script was rolled back , I have to choose a more cautious approach .

Each project has its own pipeline , Project cost , Just like the example I just gave , The time of the testing department is usually compressed , Especially near the launch or version update , Any behavior that affects the test process may cause the project to fail to be accepted normally , And then cause the project to delay or suffer Bug Online and other serious consequences , This is absolutely not allowed for a mature project that has been launched .

Optimization is not just a technical activity

My predecessor made a version of optimization before leaving , As a result, there are nearly twice as many game resource files . Finally, the game download time is too long , Rejected by Apple . So the first task I took over , It turned out to be working overtime all night and seriously retreating from the optimization code of our predecessors , That taste can be imagined how absurd and helpless .

So optimization is a trade-off , Timing , For people , It varies from project to project , Analyze and clarify the key points of optimization , What will be sacrificed at this price , It is reasonable to optimize after confirming . If the analysis is not clear, blindly optimize , You may not get the effect or make negative optimization , At the same time, he also buried many uncertain pits in his future .

For example, games focus on combat experience , Then fighting without Caton is the most needed part , You may preload , Object pool , Special effect grading and other technologies to ensure CPU and GPU Fluency of . Although increasing the memory may cause the low-end machine to flash back , But it's always better than fighting. It's better for every machine to get stuck one by one . After ensuring the game experience , Then slowly find a way to optimize the memory occupation . In fact, the best way to exchange time and space is to transfer the cost to external storage . Some are right IO Time consumption insensitive , For example, some UI Data read by logic , You can put it in an external directory , Instead of putting it in memory . It's also good to choose a suitable degree according to the final experience . Of course, if this leads to too many external storage files , And the project has complex logic , That will also bring other problems , Still have to make a choice . Of course, if your core gameplay is not so hard , It's like making a SLG, Require the game to iterate over new content , Make late profits , Then you should be more cautious about the design of the framework , Maintainability and scalability of the framework , Rapid response has become the focus of a project . So no matter time or space , Make concessions to the architecture without affecting the experience , Make the architecture more flexible and easy to use .

So the optimization problem is very big , In fact, the most important thing is to figure out the goal , What do you want to optimize . First determine the focus of the project , Then choose the optimization direction , Then analyze the bottleneck of the project through appropriate tools , Then choose the appropriate method to optimize . Of course, we should remember that sentence in the process ：

Control complexity , Make a choice

Search for

Copy