Add an example for ray queries and acceleration structures

[atynagano]

Fixes #2308.

The code is mostly from triangle-v1_3 and lacks compatibility check, comments, and may contains irregular coding styles.

[Rua]

Are you still working on this?

[atynagano]

I am not confident about the quality, but should I make it Ready for review?

[marc0246]

Could you please remove all the comments that aren't related to ray queries? For consitency with the other examples. Also, please run rustfmt on the code.

[Rua]

At first glance it looks ok, but acceleration structures aren't supported on my device yet (still waiting on Mesa 23.2) so I can't test it. The code should not crash when it's not supported, so that's a TODO for this PR still.

I notice that when you build an acceleration structure, you immediately submit a command buffer and then wait for it to finish. That doesn't seem very efficient.

[atynagano]

@Rua Thanks for your review.

I notice that when you build an acceleration structure, you immediately submit a command buffer and then wait for it to finish. That doesn't seem very efficient.

I know, but BLAS, TLAS, and pipeline have dependencies, and at least these cannot be executed asynchronously. I have referred to the following implementation, where they are executed synchronously.

https://github.com/KhronosGroup/Vulkan-Samples/blob/5e4e3d20b3a86f7814e026042341cf34ef530edd/framework/core/acceleration_structure.cpp#L218
https://github.com/KhronosGroup/Vulkan-Samples/blob/5e4e3d20b3a86f7814e026042341cf34ef530edd/framework/core/device.cpp#L621

[Rua]

But is it necessary to wait for the submission? That means the whole execution stalls until the fence is signalled. A semaphore, or just putting them all in the same command buffer, seems less extreme.

[atynagano]

I got it, but I don't know how to realize it in vulkano.