Using Radeon to simulate reflections takes up GPU space but has no effect

[ZheYinX]

I am using the C API version 4.8.1. When I attempt to use Radeon for reflection, I notice GPU usage increases, but the reflection itself does not seem to have any effect. However, when I switch back to Default or other options, the reflection works as expected.

Additionally, I am using Direct occlusion, and it works correctly regardless of the SceneType setting—whether it's set to Default, Radeon, or any other option.

During baking, GPU usage can reach 100%, but the baked audio still shows no reflection effect. I'm unsure where the issue might be. I have set both the simulator and the Scene's SceneType to Radeon, and I have also configured OpenCL and Radeon devices in SceneSettings as well as the simulator settings.

Another problem I'm facing is that I'm using SteamAudio through a C# binding library. Directly calling OpenCLListCreate in C# causes a crash. However, if I create a C++ wrapper and call an arbitrary invalid method from SteamAudio—even something like sizeof(IPLContextSettings)—before creating the list in C#, it no longer crashes and the device is created successfully. I'm not sure why this happens. I have spent a lot of time debugging but have not found the cause; structure sizes and offsets seem correct.

If any code examples are relevant, I can include them in a code block (using ```).

Thank you for your help!

[ZheYinX]

Okay, I've found the reason. I set the geometry material too extreme, which further illustrates the inconsistency between Radeon's reflection behavior and CPU's reflection behavior. I feel that CPU's reflection effect will be better, but I don't know if Radeon's effect can be improved in the future.
Although this issue has been resolved, my doubts about the crash of OpenCLList creation have not been resolved. Does SteamAudio have initialization code outside of IPLContextCreate that can only be executed in C++?