Black grid on camera frame when displaying point cloud
I am developing an application that receives and display a point cloud. When I try to render a point cloud (both with normal mesh or MaterialPropertyBlock and GPU) I get what it seems to be a black transparent canvas that follows my head. This is very annoying. I tried to change materials and shaders, to disable shadows, but I didn't find a way to get rid of this problem.
Oculus Quest - huge render textures in rendering profiler
Rendering empty scene (solid color) in Unity 2019.1.10f takes Quest's GPU around 10ms. After further profiling I discovered that there are 4 render textures (I'm guessing color and depth for each eye) and that they take almost 300MB of space. Snapdragon Profiler showed that texture memory read is often up to 6.25 MB/s. Profiling the game with Renderdoc didn't show any huge textures, eye textures were taking around 13.07MB (1216x1344) What could be the cause of this?(Android) Unity got Poor CPU Performance compared with Native apk
Hello everyone, my team have been contributing to develop " view dependent media player ", which using a single thread decode 2048 * 1536 with android-media-codec (we can call it BASE TEX) and four thread decode 512 * 256 * 24 split blocks for viewer directly look at with ffmpeg (we can call it HD TEX), very similar with VR_5K_PLAYER. The above main framework make the cpu usage more heavy compare with gpu. Cpu need decode 2048 *1536 totally and upload YUV from memory to video memory every frame, while gpu only decode 2048 *1536. After a few test with Samsung S8 & Quest, i found that, s8 & Quest got different performance which s8 got a high fps because of Quest has less cpu avaliability compare with S8. ( PS: S8 & Quest all equiped with snapdragon 835. as far as i know, the 835 chip contain 8 core, 4 core of all 8 core are big core responsible for heavy work ) To confirm my point of view , i made a test proj using unity & the latest gear sdk, add heavy cpu work for some thread public class NewBehaviourScript : MonoBehaviour { // Use this for initialization void Start () { for (int i = 0; i < 8; i++) { ThreadStart method = () => threadWork(); Thread thread = new Thread(method); thread.Start(); } void Update() { long sum = 0; for (int i = 0; i < 1000000; ++i) sum = (int)((sum + i) * 2.0f / 2.0f); } void threadWork() { while (true) { long sum = 0; for (int i = 0; i < 100000000; ++i) sum = (int)((sum + i) * 2.0f / 2.0f); Debug.LogFormat("TestThread End name:{0} curr thread id:{1} cur timeStamp:{2}", sum, Thread.CurrentThread.ManagedThreadId, DateTime.Now.ToString()); } } } the enclosed python file use "adb shell cat /proc/stat" to calculate 8 core usage. the above result is S8, the last four core are big core usage, while 3 big core execute full load. the above result is Quest, the last four core are big core usage, while 2 big core execute full load. my team also made a native android project which using gradle & android studio & without gear, loading my dynamic library through jni & execute view dependent media player. the S8 cpu usage is the below result is executed with S8 using unity & gear & same dynamic library, the two test show that native apk get a better & average cpu usage while heavy cpu work executing, compared with unity & gear apk. My question is, Is there a way for unity & gear apk to control device cpu availability rate ? why native apk performance diff with unity & gear apk? why Quest got less core to execute cpu work compared with S8?
