[wgpu-hal] Initial Vulkan ray query and acceleration structure support #3020
Conversation
|
Besides documentation and dummy implementations for the other backends, I think this is ready for review. |
…y implementation, run cargo clippy, fix wgpu/wgpu build
…me links to implementation resources.
…u into hal-acceleration-structures
|
|
||
| let geometry_info = geometry_info.build(); | ||
|
|
||
| let range = &[range][..]; |
| align_mask: if desc | ||
| .usage | ||
| .contains(crate::BufferUses::TOP_LEVEL_ACCELERATION_STRUCTURE_INPUT) | ||
| { | ||
| 16 | ||
| } else { | ||
| req.alignment | ||
| } - 1, |
There was a problem hiding this comment.
Let's pull the if statement into a variable.
| vk::WriteDescriptorSetAccelerationStructureKHR::builder() | ||
| .acceleration_structures(&raw_acceleration_structures[raw_start..]); | ||
|
|
||
| // todo: Dereference the struct to get around lifetime issues. Safe as long as we never resize |
There was a problem hiding this comment.
If we need to do this, let's add an assertion that we always have room on push.
| instance_custom_index_and_mask: pack_24_8(0, 0xff), | ||
| instance_shader_binding_table_record_offset_and_flags: pack_24_8(0, 0), | ||
| acceleration_structure_reference: unsafe { | ||
| device.get_acceleration_structure_device_address(&blas) |
There was a problem hiding this comment.
This is the kind of thing we'd hide at the wgpu-core level?
| const ACCELERATION_STRUCTURE_SCRATCH = 1 << 10; | ||
| const BOTTOM_LEVEL_ACCELERATION_STRUCTURE_INPUT = 1 << 11; | ||
| const TOP_LEVEL_ACCELERATION_STRUCTURE_INPUT = 1 << 12; |
There was a problem hiding this comment.
These probably need to be added to either INCLUSIVE or EXCLUSIVE in some way?
|
Poke from triage |
|
This PR is currently on-pause because I don't have the time to work on it along with everything else I'm doing. I don't think it'll take much to get it to a mergeable state and if someone else wants to work on that then I'm happy to review their changes. I wish there was a way to indicate this on github, lots of people (understandably) seem to be afraid to touch someone else's PR if it's still open |
|
@cwfitzgerald is there a list of TODOs left on this before being mergeable? I'd like to pick this up |
|
@seabassjh love to have this be picked up - the main ones are the comments on the PR |
I imagine there's some work to be done to resolve #3044 etc as well |
|
I would like to help out. |
|
@daniel-keitel if you link me your changes I can review them and merge them into this branch |
|
@expenses I created a pull request on your fork. |
|
Hi @daniel-keitel, sorry I missed this. It might be better if you created a new PR on wgpu and then the discussion moved over to there. |
|
Closing on favor of 3507 |
Checklist
cargo clippy.RUSTFLAGS=--cfg=web_sys_unstable_apis cargo clippy --target wasm32-unknown-unknownif applicable.Connections
#1040
Description
Still a work in progress, but I can ray-trace instanced triangles now:
Needs a cleanup and support for acceleration structure updates.
Testing
Explain how this change is tested.