Error: Out of memory Error: vkAllocateMemory failed with VK_ERROR_OUT_OF_DEVICE_MEMORY

Nobody has claimed this yet.

Hello @maierfelix

I use procedural geometry and add 4 million instances to top level acceleration structure(bottom level acceleration structure only has 1 aabb), then when I run the demo, it error:
```js
Error: Out of memory Error: vkAllocateMemory failed with VK_ERROR_OUT_OF_DEVICE_MEMORY
at CheckVkOOMThenSuccessImpl (../../src/dawn_native/vulkan/VulkanError.cpp:94)
at AllocateResourceHeap (../../src/dawn_native/vulkan/ResourceMemoryAllocatorVk.cpp:87)
at Allocate (../../src/dawn_native/vulkan/ResourceMemoryAllocatorVk.cpp:140)
at Initialize (../../src/dawn_native/vulkan/BufferVk.cpp:171)
at Create (../../src/dawn_native/vulkan/BufferVk.cpp:125)
at device._onErrorCallback (D:\Github\WebGPU-Example\node_modules\wonder-webgpu\index.js:57:21)
at _buildContainers (file:///D:/Github/WebGPU-Example/packages/arc_ray_tracing_pipeline/src/pipeline/jobs/init/InitRayTracingPassJob.js:148:36)
at exec (file:///D:/Github/WebGPU-Example/packages/arc_ray_tracing_pipeline/src/pipeline/jobs/init/InitRayTracingPassJob.js:179:29)
at exec (file:///D:/Github/WebGPU-Example/packages/arc_ray_tracing_pipeline/src/pipeline/InitPipeline.js:19:13)
at async _main (file:///D:/Github/WebGPU-Example/packages/arc_ray_tracing_pipeline/src/index.js:52:13)
```

I want to add at least >10 million instances to top level acceleration structure! How to handle the error?

Thank you very much!

Start by reproducing the reported demo and follow the allocation stack through InitRayTracingPassJob.js, BufferVk.cpp, and ResourceMemoryAllocatorVk.cpp. Check how the four-million-instance acceleration structure reaches vkAllocateMemory and document whether the requested instance count is supported or what limit and failure behavior should be expected.

Written by the indexing model from the issue text.