Renderdoc Meta Fork not providing performance counters for compute shaders
Hello! As the title says, I have been having issues profiling compute shaders. I am doing and investigation on VR volume rendering on the Meta Quest 2; and the performance counters on the Renderdoc Meta Fork have been fundamental on my work. But, right now, the only information that is provided on the counter are the millisencos for each pass, only when taking about compute shaders (the render passes present all the data as expected). At first I thought that it is as expected for some fields to present missing values, since we are not using the raster pipeline, but the texture pipelines are also reporting 0 fetch stall & 0 bytes readed; wich is not true since this shaders perform a surface nets algortihm. Am I missing something? Has anyone encoountered this issue? Also, it is not a dead-or-live situation, but the shaders run very fast; and would love to look a bit closer.Quest2 + Renderdoc Meta Fork - Injection failed : couldn't locate the launched activity....
Hi, This started happening since last week. The development APK build launches through RenderDoc OK, but while the application is loading, Renderdoc Meta Fork returns this error message: I've tried Renderdoc Meta Fork versions 39.2, 39.6,40.1,44.1,50.1 and 51.1. It seems to be quite random when it appears. The only way to fix it at the moment seems to be to keep installing various versions on the Quest2 until the error doesn't appear anymore. Sometimes the ADB and Renderdoc log/dmp files get updated during a crash, examples are shared here: https://drive.google.com/drive/folders/1-z4-VxMAwEnXskJ6hnngPlxAIaIWNVKC
OVR Metrics tool showing GPU stats as 0
I have been using the Metrics tool as my #1 tool for in-app profiling and testing, but now the one feature I needed no longer works. I need to the frag/vert/frag%/vert% to guide me when I'm optimizing and all of those all show 0 now. Is this maybe an OS regression issue? I wasn't having this issue a few days ago. I have tried downloading all the older versions to see if it's the OVR metrics tool, all of them behaved the same. I have tried multiple headsets and they behave the same.
Setting GPU level on Quest 2
I'm trying to profile a Unity app on the Quest 2, but I'm having trouble setting the GPU level. Here is what I've been trying to do: I'm checking the current GPU level using adb logcat -s VrApi (under CPU4/GPU). It seems to hover around 2 or 3. I tried using adb shell getprop debug.oculus.gpuLevel, but by default this returns nothing. I assume this property doesn't refer to the actual GPU Level but the enforced GPU Level, so having it empty means dynamic GPU level. In my Unity project, I'm using: OVRManager.gpuLevel = 2; to try and fix the GPU level. This seems to be ignored by the Quest (logcat still reports GPU level 3 sometimes). Moreover, using adb shell getprop still returns nothing. Shouldn't that one pick up the fixed GPU Level now? Alternatively, I tried using adb shell to fix the GPU level at runtime using: adb shell setprop debug.oculus.gpuLevel 2 When I do that, getprop does pick up on the value I've set, but the GPU level reported by logcat still fluctuates sometimes. Any idea what's going on there? Should any of these be working, or am I missing a step? We just tried the adb shell solution on a colleague's Quest 1 and it worked out of the box (though he was running a different app). Having no control over the GPU level makes profiling a bit of a pain...Profiling app RAM usage on the Quest
Hey guys, I am getting inconsistent stats from Logcat and Snapdragon Profiler. I am profiling the RAM usage of my application made with Unreal Engnine 4.24 and latest Oculus SDK (from the oculus branch of unreal engine). The issue is that according to snap dragon my app is using around 1.5 Gb and according to Logcat 227 mb. (I have already posted this on the Snapdragon developer forum) Who do I trust? And how do you guys profile your RAM usage?
