refresh: stack al día (vello 0.7 / wgpu 27 / parley 0.6) + motor 3D voxel
Re-sincroniza las fuentes desde el monorepo (estaba en vello 0.5/wgpu 24 y con la estructura vieja de eventloop) y suma el 3D: - bump del workspace a vello 0.7 / wgpu 27 / parley 0.6, + accesskit 0.24 / accesskit_winit 0.33 / vello_hybrid 0.0.9. - nuevos crates: llimphi-3d (voxels ray-march + mallas en un depth compartido, montable dentro de un View 2D vía set_viewport+scissor) y llimphi-voxel (world-gen, personajes, director de escenas) + shared/foreign-vox (puente .vox). - README: sección "Not just 2D — a 3D voxel engine" + GIF (docs/llimphi_voxel.gif). - excluido modules/allichay (arrastra deps fuera del alcance del front-door). - cargo check --workspace: verde. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Sergio
@@ -0,0 +1,314 @@
|
||||
//! `Renderer3d` — pipeline wgpu mínimo que dibuja geometría 3D indexada con
|
||||
//! test de profundidad sobre la textura intermedia del frame de Llimphi.
|
||||
//!
|
||||
//! La firma de [`Renderer3d::render`] es la que pide la closure de
|
||||
//! `View::gpu_paint_with` (`device, queue, encoder, target_view, (w, h)`), más
|
||||
//! la cámara — así un nodo 3D entra en el árbol `View<Msg>` sin tocar el
|
||||
//! runtime. Mantiene su **propio depth buffer** (recreado al cambiar de
|
||||
//! tamaño); el color se compone con `LoadOp::Load` para preservar la UI vello
|
||||
//! que ya está debajo.
|
||||
|
||||
use glam::Mat4;
|
||||
|
||||
use crate::camera::Camera3d;
|
||||
use crate::mesh::{cube, Vertex3d};
|
||||
use crate::scene::{ensure_depth, DepthBuffer, DEPTH_FORMAT};
|
||||
|
||||
/// Renderer de **mallas** indexadas (por defecto un cubo) visto desde una
|
||||
/// [`Camera3d`]. Cachea pipeline, buffers de geometría, uniform y (para el
|
||||
/// camino standalone) un depth propio. En [`Scene3d`](crate::Scene3d) comparte
|
||||
/// el depth con el pase de voxels para ocluirse mutuamente.
|
||||
///
|
||||
/// `model` ubica la malla en el mundo (default identidad): `mvp = view_proj ·
|
||||
/// model`, así una misma malla se instancia/posiciona sin reconstruir buffers.
|
||||
pub struct Renderer3d {
|
||||
pipeline: wgpu::RenderPipeline,
|
||||
vbuf: wgpu::Buffer,
|
||||
ibuf: wgpu::Buffer,
|
||||
index_count: u32,
|
||||
ubuf: wgpu::Buffer,
|
||||
bind_group: wgpu::BindGroup,
|
||||
model: Mat4,
|
||||
depth: Option<DepthBuffer>,
|
||||
}
|
||||
|
||||
impl Renderer3d {
|
||||
/// Crea el renderer para un `color_format` dado (el de la textura
|
||||
/// intermedia del frame — `Rgba8Unorm` en headless, el de la surface en
|
||||
/// vivo). Arranca con el cubo de prueba cargado.
|
||||
pub fn new(device: &wgpu::Device, color_format: wgpu::TextureFormat) -> Self {
|
||||
let (verts, indices) = cube();
|
||||
Self::with_mesh(device, color_format, &verts, &indices)
|
||||
}
|
||||
|
||||
/// Igual que [`Self::new`] pero con una malla arbitraria.
|
||||
pub fn with_mesh(
|
||||
device: &wgpu::Device,
|
||||
color_format: wgpu::TextureFormat,
|
||||
verts: &[Vertex3d],
|
||||
indices: &[u16],
|
||||
) -> Self {
|
||||
let shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {
|
||||
label: Some("llimphi-3d-shader"),
|
||||
source: wgpu::ShaderSource::Wgsl(WGSL.into()),
|
||||
});
|
||||
|
||||
let bind_layout = device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
|
||||
label: Some("llimphi-3d-bgl"),
|
||||
entries: &[wgpu::BindGroupLayoutEntry {
|
||||
binding: 0,
|
||||
visibility: wgpu::ShaderStages::VERTEX,
|
||||
ty: wgpu::BindingType::Buffer {
|
||||
ty: wgpu::BufferBindingType::Uniform,
|
||||
has_dynamic_offset: false,
|
||||
min_binding_size: None,
|
||||
},
|
||||
count: None,
|
||||
}],
|
||||
});
|
||||
|
||||
let pipeline_layout = device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
|
||||
label: Some("llimphi-3d-pl"),
|
||||
bind_group_layouts: &[&bind_layout],
|
||||
push_constant_ranges: &[],
|
||||
});
|
||||
|
||||
let pipeline = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
|
||||
label: Some("llimphi-3d-pipeline"),
|
||||
layout: Some(&pipeline_layout),
|
||||
vertex: wgpu::VertexState {
|
||||
module: &shader,
|
||||
entry_point: Some("vs"),
|
||||
compilation_options: Default::default(),
|
||||
buffers: &[wgpu::VertexBufferLayout {
|
||||
array_stride: Vertex3d::SIZE as u64,
|
||||
step_mode: wgpu::VertexStepMode::Vertex,
|
||||
attributes: &[
|
||||
wgpu::VertexAttribute {
|
||||
format: wgpu::VertexFormat::Float32x3,
|
||||
offset: 0,
|
||||
shader_location: 0,
|
||||
},
|
||||
wgpu::VertexAttribute {
|
||||
format: wgpu::VertexFormat::Float32x3,
|
||||
offset: 12,
|
||||
shader_location: 1,
|
||||
},
|
||||
],
|
||||
}],
|
||||
},
|
||||
primitive: wgpu::PrimitiveState {
|
||||
topology: wgpu::PrimitiveTopology::TriangleList,
|
||||
strip_index_format: None,
|
||||
front_face: wgpu::FrontFace::Ccw,
|
||||
// M0 sin cull: el depth test ya resuelve la oclusión y nos
|
||||
// ahorra bugs de winding al sumar mallas. El cull entra en M1+.
|
||||
cull_mode: None,
|
||||
unclipped_depth: false,
|
||||
polygon_mode: wgpu::PolygonMode::Fill,
|
||||
conservative: false,
|
||||
},
|
||||
depth_stencil: Some(wgpu::DepthStencilState {
|
||||
format: DEPTH_FORMAT,
|
||||
depth_write_enabled: true,
|
||||
depth_compare: wgpu::CompareFunction::Less,
|
||||
stencil: wgpu::StencilState::default(),
|
||||
bias: wgpu::DepthBiasState::default(),
|
||||
}),
|
||||
multisample: wgpu::MultisampleState::default(),
|
||||
fragment: Some(wgpu::FragmentState {
|
||||
module: &shader,
|
||||
entry_point: Some("fs"),
|
||||
compilation_options: Default::default(),
|
||||
targets: &[Some(wgpu::ColorTargetState {
|
||||
format: color_format,
|
||||
// Opaco: el cubo reemplaza el fondo vello donde lo cubre.
|
||||
blend: None,
|
||||
write_mask: wgpu::ColorWrites::ALL,
|
||||
})],
|
||||
}),
|
||||
multiview: None,
|
||||
cache: None,
|
||||
});
|
||||
|
||||
// Geometría → buffers (idiom `to_ne_bytes`, sin bytemuck).
|
||||
let mut vbytes = Vec::with_capacity(verts.len() * Vertex3d::SIZE);
|
||||
for v in verts {
|
||||
v.write_to(&mut vbytes);
|
||||
}
|
||||
let vbuf = create_buffer_init(device, "llimphi-3d-vbuf", wgpu::BufferUsages::VERTEX, &vbytes);
|
||||
|
||||
let mut ibytes = Vec::with_capacity(indices.len() * 2);
|
||||
for &i in indices {
|
||||
ibytes.extend_from_slice(&i.to_ne_bytes());
|
||||
}
|
||||
let ibuf = create_buffer_init(device, "llimphi-3d-ibuf", wgpu::BufferUsages::INDEX, &ibytes);
|
||||
|
||||
let ubuf = device.create_buffer(&wgpu::BufferDescriptor {
|
||||
label: Some("llimphi-3d-ubuf"),
|
||||
size: 64, // una mat4x4<f32>
|
||||
usage: wgpu::BufferUsages::UNIFORM | wgpu::BufferUsages::COPY_DST,
|
||||
mapped_at_creation: false,
|
||||
});
|
||||
|
||||
let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
|
||||
label: Some("llimphi-3d-bg"),
|
||||
layout: &bind_layout,
|
||||
entries: &[wgpu::BindGroupEntry {
|
||||
binding: 0,
|
||||
resource: ubuf.as_entire_binding(),
|
||||
}],
|
||||
});
|
||||
|
||||
Self {
|
||||
pipeline,
|
||||
vbuf,
|
||||
ibuf,
|
||||
index_count: indices.len() as u32,
|
||||
ubuf,
|
||||
bind_group,
|
||||
model: Mat4::IDENTITY,
|
||||
depth: None,
|
||||
}
|
||||
}
|
||||
|
||||
/// Ubica la malla en el mundo (`mvp = view_proj · model`). Default identidad.
|
||||
pub fn set_model(&mut self, model: Mat4) {
|
||||
self.model = model;
|
||||
}
|
||||
|
||||
/// Reemplaza la geometría (recrea los buffers de vértices/índices). Pensado
|
||||
/// para mallas que cambian cada frame — p.ej. un **muñeco articulado** cuya
|
||||
/// pose se rehornea en CPU (limbos rotados) y se vuelve a subir. Las mallas
|
||||
/// son chicas (decenas-cientos de vértices), así que recrear los buffers por
|
||||
/// frame es despreciable; el pipeline/uniform/bind-group se conservan.
|
||||
pub fn set_geometry(&mut self, device: &wgpu::Device, verts: &[Vertex3d], indices: &[u16]) {
|
||||
let mut vbytes = Vec::with_capacity(verts.len() * Vertex3d::SIZE);
|
||||
for v in verts {
|
||||
v.write_to(&mut vbytes);
|
||||
}
|
||||
self.vbuf =
|
||||
create_buffer_init(device, "llimphi-3d-vbuf", wgpu::BufferUsages::VERTEX, &vbytes);
|
||||
|
||||
let mut ibytes = Vec::with_capacity(indices.len() * 2);
|
||||
for &i in indices {
|
||||
ibytes.extend_from_slice(&i.to_ne_bytes());
|
||||
}
|
||||
self.ibuf =
|
||||
create_buffer_init(device, "llimphi-3d-ibuf", wgpu::BufferUsages::INDEX, &ibytes);
|
||||
self.index_count = indices.len() as u32;
|
||||
}
|
||||
|
||||
/// Sube el uniform del frame (`mvp = view_proj · model`). Lo llama
|
||||
/// [`Self::render`] y [`Scene3d`](crate::Scene3d). `aspect` = w/h.
|
||||
pub fn upload(&self, queue: &wgpu::Queue, aspect: f32, camera: &Camera3d) {
|
||||
let mvp = camera.view_proj(aspect) * self.model;
|
||||
// glam es column-major; el shader WGSL espera column-major → upload tal cual.
|
||||
let mut ubytes = Vec::with_capacity(64);
|
||||
for v in mvp.to_cols_array() {
|
||||
ubytes.extend_from_slice(&v.to_ne_bytes());
|
||||
}
|
||||
queue.write_buffer(&self.ubuf, 0, &ubytes);
|
||||
}
|
||||
|
||||
/// Dibuja la malla indexada en un pase **ya abierto** (color + depth). Lo usa
|
||||
/// [`Scene3d`](crate::Scene3d) para compartir el pase con los voxels.
|
||||
/// Requiere `upload` previo en el mismo frame.
|
||||
pub fn draw<'a>(&'a self, pass: &mut wgpu::RenderPass<'a>) {
|
||||
pass.set_pipeline(&self.pipeline);
|
||||
pass.set_bind_group(0, &self.bind_group, &[]);
|
||||
pass.set_vertex_buffer(0, self.vbuf.slice(..));
|
||||
pass.set_index_buffer(self.ibuf.slice(..), wgpu::IndexFormat::Uint16);
|
||||
pass.draw_indexed(0..self.index_count, 0, 0..1);
|
||||
}
|
||||
|
||||
/// Dibuja la malla sola sobre `target` (camino standalone, depth propio).
|
||||
/// Firma compatible con `View::gpu_paint_with`; color preservado
|
||||
/// (`LoadOp::Load`), depth propio limpiado cada frame.
|
||||
pub fn render(
|
||||
&mut self,
|
||||
device: &wgpu::Device,
|
||||
queue: &wgpu::Queue,
|
||||
encoder: &mut wgpu::CommandEncoder,
|
||||
target: &wgpu::TextureView,
|
||||
(w, h): (u32, u32),
|
||||
camera: &Camera3d,
|
||||
) {
|
||||
if w == 0 || h == 0 {
|
||||
return;
|
||||
}
|
||||
self.upload(queue, w as f32 / h as f32, camera);
|
||||
ensure_depth(&mut self.depth, device, w, h);
|
||||
let depth_view = &self.depth.as_ref().unwrap().view;
|
||||
|
||||
let mut pass = encoder.begin_render_pass(&wgpu::RenderPassDescriptor {
|
||||
label: Some("llimphi-3d-pass"),
|
||||
color_attachments: &[Some(wgpu::RenderPassColorAttachment {
|
||||
view: target,
|
||||
resolve_target: None,
|
||||
depth_slice: None,
|
||||
ops: wgpu::Operations {
|
||||
load: wgpu::LoadOp::Load,
|
||||
store: wgpu::StoreOp::Store,
|
||||
},
|
||||
})],
|
||||
depth_stencil_attachment: Some(wgpu::RenderPassDepthStencilAttachment {
|
||||
view: depth_view,
|
||||
depth_ops: Some(wgpu::Operations {
|
||||
load: wgpu::LoadOp::Clear(1.0),
|
||||
store: wgpu::StoreOp::Store,
|
||||
}),
|
||||
stencil_ops: None,
|
||||
}),
|
||||
timestamp_writes: None,
|
||||
occlusion_query_set: None,
|
||||
});
|
||||
self.draw(&mut pass);
|
||||
}
|
||||
}
|
||||
|
||||
/// Crea un buffer ya inicializado con `data` (sin `wgpu::util::DeviceExt`, para
|
||||
/// no arrastrar la feature `util`): `mapped_at_creation` + copia + `unmap`.
|
||||
fn create_buffer_init(
|
||||
device: &wgpu::Device,
|
||||
label: &str,
|
||||
usage: wgpu::BufferUsages,
|
||||
data: &[u8],
|
||||
) -> wgpu::Buffer {
|
||||
let buf = device.create_buffer(&wgpu::BufferDescriptor {
|
||||
label: Some(label),
|
||||
size: data.len() as u64,
|
||||
usage,
|
||||
mapped_at_creation: true,
|
||||
});
|
||||
buf.slice(..).get_mapped_range_mut().copy_from_slice(data);
|
||||
buf.unmap();
|
||||
buf
|
||||
}
|
||||
|
||||
const WGSL: &str = r#"
|
||||
struct Uniforms { mvp: mat4x4<f32> };
|
||||
@group(0) @binding(0) var<uniform> u: Uniforms;
|
||||
|
||||
struct VIn {
|
||||
@location(0) pos: vec3<f32>,
|
||||
@location(1) color: vec3<f32>,
|
||||
};
|
||||
struct VOut {
|
||||
@builtin(position) clip: vec4<f32>,
|
||||
@location(0) color: vec3<f32>,
|
||||
};
|
||||
|
||||
@vertex
|
||||
fn vs(in: VIn) -> VOut {
|
||||
var out: VOut;
|
||||
out.clip = u.mvp * vec4<f32>(in.pos, 1.0);
|
||||
out.color = in.color;
|
||||
return out;
|
||||
}
|
||||
|
||||
@fragment
|
||||
fn fs(in: VOut) -> @location(0) vec4<f32> {
|
||||
return vec4<f32>(in.color, 1.0);
|
||||
}
|
||||
"#;
|
||||
Reference in New Issue
Block a user