Vulkan is not a "panacea"“

The previous articles , We introduced a new generation of graphics API-Vulkan Advantages and characteristics of . Because of its fine control and performance characteristics , Some little friends have raised questions “ Whether the existing OpenGL The project moved to Vulkan On ？”,“Vulkan How is it that the effect is not as good as expected ？”. Today, let's talk about this topic .

The first thing I want to say is Vulkan Not all rendering problems can be solved by replacing graphics API It can be easily solved , There may even be negative benefits and costs .

Scene one Business performance problems

The problems of their own business scenarios have not been effectively solved , It's a drop in the bucket just to rely on the promotion of the bottom . All ideas to solve performance are to measure first , Then optimize the existing problems of the project according to the measurement results , Then through the underlying foundation and general capabilities . And most of the scenes are their own problems that have not been well solved .

Scene two Problems in non rendered scenes

The current performance bottleneck is not rendering related , It may be data, business logic or even network .

Scene three Low latency and less jamming

If your application is concerned about tiny catons or frame rate jitters ,Vulkan You can explicitly control when time-consuming operations occur during scene rendering . This benefits from Vulkan Fine control and design philosophy , Compared with OpenGL Drivers have more advantages in managing state and resources through prediction .

Scene 4   Multithreaded rendering

If the bottleneck of program performance is CPU Graphics related parts on , that Vulkan There is likely to be an opportunity to improve its performance , because Vulkan It naturally supports multithreading , Can give full play to CPU The ability of , Compared with OpenGL Single threaded rendering of has more advantages . Or for the performance of a platform that wants to squeeze out relatively limited computing resources , that Vulkan Allowing the program to directly allocate and manage all resources may also help performance .

Scene five Performance bottleneck GPU

  This is a scene that is easy to cause controversy and misunderstanding , I personally understand the essence of graphics API Is with the GPU A specification that drives communication ,GPU The driver is CPU And GPU The medium of communication , All of these （ graphics API、 drive ） It's all implemented in CPU End , Instead of GPU End .

therefore , Different graphics API It cannot be directly promoted GPU Own work efficiency and improve its load , Can only improve CPU End and GPU Efficiency of end-to-end communication , Collaborative approach .

Although it cannot directly affect GPU Performance of , But different graphics API Yes GPU The granularity of the control ability is different ,Vulkan Far ratio OpenGL Handled carefully , Can also be used （ customized ） added GPU End function , So in some scenes, replace with Vulkan It can provide us with more control means , Is likely to help improve GPU Efficiency of execution .

Scene six Ecology and cost

Whether it's technology ecology or human cost ,Vulkan Are relatively at a disadvantage , So we must consider this when doing project migration , You can't follow blindly .

Of course, there are thousands of business scenarios , Developers need to consider many aspects to seek the greatest ROI.

attach ：

Vulkan- Practical analysis

Vulkan- Performance and fine control