Meshlet BVH Culling #19318
Meshlet BVH Culling #19318
Conversation
atlv24
added
C-Feature
| // let world_sphere_radius = lod_sphere.w * world_scale; | ||
| // let norm_error = simplification_error / world_sphere_radius * 0.25; | ||
| // return norm_error * view.viewport.w < 1.0; | ||
| let world_error = simplification_error * world_scale; | ||
| let proj = projection[1][1]; | ||
| let height = 2.0 / proj; | ||
| let norm_error = world_error / height; | ||
| return norm_error * view.viewport.w < 1.0; |
There was a problem hiding this comment.
Unsure how correct my change (from the commented code) is, though it seems to work.
There was a problem hiding this comment.
Ok did my first pass of this.
A lot of it I just skimmed because there's way too much code for me to verify line by line, so I'm just going to trust a lot of the BVH traversal stuff and resource handling is correct.
Here's what I still need to verify:
- from_mesh() code
- Occlusion culling changes
- LOD selection changes
- Try it out and test perf (my current computer doesn't support 64bit atomics unfortunately)
| @@ -233,6 +233,7 @@ pub fn extract_meshlet_mesh_entities( | |||
| } | |||
|
|
|||
| // Iterate over every instance | |||
| // TODO: Switch to change events to not upload every instance every frame. | |||
There was a problem hiding this comment.
Yeah better instance handling on the CPU is a big todo. I know @tychedelia is also gonna look into caching material specialization for meshlet instances as well.
| #import bevy_render::maths::affine3_to_square | ||
|
|
||
| // https://github.com/zeux/meshoptimizer/blob/1e48e96c7e8059321de492865165e9ef071bffba/demo/nanite.cpp#L115 | ||
| fn lod_error_is_imperceptible(lod_sphere: vec4<f32>, simplification_error: f32, instance_id: u32) -> bool { |
There was a problem hiding this comment.
TODO: I need to review this carefully
| } | ||
| }; | ||
| match &mut resource_manager.cluster_meshlet_ids { | ||
| instance_manager |
There was a problem hiding this comment.
Did you test that this is equally as fast as upload_storage_buffer?
There was a problem hiding this comment.
If I remember correctly, upload_storage_buffer did the same thing internally so I just got rid of it.
There was a problem hiding this comment.
I think I might've added that after writing upload_storage_buffer 😅
| @@ -19,11 +23,13 @@ use meshopt::{ | |||
| use metis::{option::Opt, Graph}; | |||
| use smallvec::SmallVec; | |||
| use thiserror::Error; | |||
| use tracing::debug_span; | |||
|
|
|||
There was a problem hiding this comment.
TODO: I need to review this whole file, haven't looked at it yet
What was wrong with each of these? |
|
I don't remember what was wrong with them anymore. I do remember the symptoms though: |
|
You added a new example but didn't add metadata for it. Please update the root Cargo.toml file. |
crates/bevy_pbr/src/meshlet/asset.rs
Outdated
| writer.flush()?; | ||
| writer.flush()?; | ||
| writer.flush()?; | ||
| writer.flush()?; |
atlv24
force-pushed
the
bvh-cull-fixed
branch
2 times, most recently
from
9a3d12c to
82df1b1
Compare
Objective
Solution
Testing