New features mail GPU counter module adds GPU primitive processing and GPU shader cycles

Mail GPU Counter modular Since its launch, it has been widely praised , There have been many items checked GPU Coloring and bandwidth information , Locate the GPU High pressure scenarios and performance bottlenecks . be based on GPU There is a growing need to optimize ,UWA Several important parameters have been added in the evaluation report of the new version , So that the developer can have a better understanding of the project GPU Pressure for a more comprehensive understanding and efficient positioning .

From now on , Use UWA SDK 2.4.3 Version in progress GOT Online Overview Mode testing , You can Mail GPU Counter Module to view the project runtime GPU Element processing and GPU Shader Cycles situation .

One 、GPU Graphic processing

Rendered faces are generated GPU One of the important factors of stress , We can go through Overview In mode Triangle Indicators to view and analyze which pictures have more rendering surfaces .

Too many render faces , On the one hand, the model may be too complex , In general, you can go through LOD、HLOD And other commonly used techniques to simplify long-distance models , Significantly reduce the rendering surface without affecting the image quality ; On the other hand , It could be the terrain 、 Large area models such as large buildings have not been properly split , There may not be many patches that lead to entry into the horizon , But submit GPU There are still many render faces .

For the second case , We can use new features “GPU Primitives ” To make a preliminary judgment .

The total number of elements ： Submitted to the GPU Total number of elements at the end , This value is basically equal to the total number of rendered patches counted by the engine .

Number of visible elements ： stay GPU After the end is cut in various ways , The triangular surface left to participate in rendering .

Visible elements do not include ： A triangle cut out of the field of view , A triangular surface cut because of orientation . therefore , stay 3D Scene , Ideally , The number of visible elements should be close to or higher than 50%（ For most models , Half of the triangles are cropped because of their orientation ）. If at some angle , The scale of visible elements is very low , The second case mentioned above is likely to exist , Thus, we can check and optimize the , From this angle , Be submitted to GPU Large area model of .

GPU Too much primitive processing will have a great impact on the bandwidth and energy consumption of the equipment , Try to eliminate at the program end , And reduce the amount sent to GPU Number of elements . Besides , The total number of elements is basically equal to the total number of rendered patches counted by the engine , Developers can also use this value to evaluate whether the number of rendered patches processed is reasonable .

Two 、GPU Shader Cycles

Shader Too much complexity is GPU One of the main reasons for high pressure , post-processing 、 Large terrain 、 Water 、 Character modeling 、 And particle effects are easy to exist Shader High complexity situations .

In order to help you check Shader Problems of excessive complexity , stay UWA These parameters are supported and detailed in the report of .Shader Cycles Number representation GPU stay Shader To dispose of Cycles Count , Show... During testing Shader Complexity situation .

Except always Shader Cycles Beyond number ,GPU Shader Cycles It also lists Shader Arithmetic Cycles、Shader Interpolator Cycles、Shader LoadStore Cycles、Shader Texture Cycles Four data . Users can quickly navigate to high Shader Complexity scenarios , Mathematical calculation 、 Vertex to pixel interpolation 、 Register read 、 Texture processing related Cycles In number , Which unit is Shader The bottleneck of high complexity , So as to make a targeted analysis of the project Shader Troubleshoot and optimize complexity .

3、 ... and 、 Other parameters and indicators are detailed

except Mail GPU Counter Outside the module , In this update ,UWA Some data items are also refined .

1、 Add temperature display
Overview The temperature module under mode hardware information is added CPU temperature 、GPU Temperature and battery temperature , To help locate the cause of the equipment heating .

2、Timeline Support to display thread name , Such as the main thread 、 Rendering threads, etc .

3、Resource New mode AssetBundle Resident information
When we check the memory problem , Will find Unity The total memory value is too high , But mainstream resources and Mono The occupancy is not high , At this time, you need to pay attention to whether it may be caused by other memory , for example AssetBundle Resident memory .

stay GOT Online Resource In mode , We can see the resource residence during the project operation . When the resident memory is large , You can reduce the number of residences by optimizing the cache policy , So as to reduce Unity Memory effect .

The above is the introduction of the new functions , I hope this support can help the R & D team locate the project more efficiently and accurately in the optimization process GPU Pain points of stress , Yield twice the result with half the effort . You can register and activate UWA You can get UWA SaaS Free trial permission of the service , Come and experience it ～