This commit is contained in:
Junkui Zhang 2025-07-13 20:17:01 +08:00
parent dbe2ce2464
commit c8ae5a3b11
2 changed files with 186 additions and 115 deletions

View file

@ -49,6 +49,7 @@ struct DirectXRenderPipelines {
shadow_pipeline: PipelineState, shadow_pipeline: PipelineState,
quad_pipeline: PipelineState, quad_pipeline: PipelineState,
paths_pipeline: PipelineState, paths_pipeline: PipelineState,
paths_indirect_draw_buffer: ID3D11Buffer,
underline_pipeline: PipelineState, underline_pipeline: PipelineState,
mono_sprites: PipelineState, mono_sprites: PipelineState,
poly_sprites: PipelineState, poly_sprites: PipelineState,
@ -61,6 +62,14 @@ struct DirectXGlobalElements {
blend_state_for_pr: ID3D11BlendState, blend_state_for_pr: ID3D11BlendState,
} }
#[repr(C)]
struct DrawInstancedIndirectArgs {
vertex_count_per_instance: u32,
instance_count: u32,
start_vertex_location: u32,
start_instance_location: u32,
}
// #[cfg(not(feature = "enable-renderdoc"))] // #[cfg(not(feature = "enable-renderdoc"))]
// struct DirectComposition { // struct DirectComposition {
// comp_device: IDCompositionDevice, // comp_device: IDCompositionDevice,
@ -122,6 +131,7 @@ impl DirectXRenderer {
[0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0],
&self.globals.blend_state, &self.globals.blend_state,
)?; )?;
println!("--> Drawing scene");
for batch in scene.batches() { for batch in scene.batches() {
match batch { match batch {
PrimitiveBatch::Shadows(shadows) => self.draw_shadows(shadows), PrimitiveBatch::Shadows(shadows) => self.draw_shadows(shadows),
@ -273,41 +283,64 @@ impl DirectXRenderer {
if paths.is_empty() { if paths.is_empty() {
return Ok(()); return Ok(());
} }
println!("Drawing {} paths", paths.len());
let mut vertices = Vec::new(); let mut vertices = Vec::new();
let mut sprites = Vec::with_capacity(paths.len()); let mut sprites = Vec::with_capacity(paths.len());
for path in paths { let mut draw_indirect_commands = Vec::with_capacity(paths.len());
let tile = &path_tiles[&path.id]; let mut start_vertex_location = 0;
let texture_view = self.atlas.get_texture_view(tile.texture_id); for (i, path) in paths.iter().enumerate() {
let origin = path.bounds.intersect(&path.content_mask.bounds).origin; draw_indirect_commands.push(DrawInstancedIndirectArgs {
let sprites = [PathSprite { vertex_count_per_instance: path.vertices.len() as u32,
bounds: Bounds { instance_count: 1,
origin: origin.map(|p| p.floor()), start_vertex_location,
size: tile.bounds.size.map(Into::into), start_instance_location: i as u32,
});
start_vertex_location += path.vertices.len() as u32;
vertices.extend(path.vertices.iter().map(|v| PathVertex {
xy_position: v.xy_position,
content_mask: ContentMask {
bounds: path.content_mask.bounds,
}, },
}));
sprites.push(PathSprite {
bounds: path.bounds,
color: path.color, color: path.color,
tile: (*tile).clone(), });
}]; }
update_buffer_capacity(
&self.pipelines.paths_pipeline, update_buffer_capacity(
std::mem::size_of::<PathSprite>(), &self.pipelines.paths_pipeline,
1, std::mem::size_of::<PathSprite>(),
&self.devices.device, sprites.len(),
) &self.devices.device,
.map(|input| update_pipeline(&mut self.pipelines.paths_pipeline, input)); )
update_buffer( .map(|input| update_pipeline(&mut self.pipelines.paths_pipeline, input));
update_buffer(
&self.devices.device_context,
&self.pipelines.paths_pipeline.buffer,
&sprites,
)?;
update_indirect_buffer(
&self.devices.device_context,
&self.pipelines.paths_indirect_draw_buffer,
&draw_indirect_commands,
)?;
prepare_indirect_draws(
&self.devices.device_context,
&self.pipelines.paths_pipeline,
&self.context.viewport,
&self.globals.global_params_buffer,
D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP,
)?;
for i in 0..paths.len() {
draw_indirect(
&self.devices.device_context, &self.devices.device_context,
&self.pipelines.paths_pipeline.buffer, &self.pipelines.paths_indirect_draw_buffer,
&sprites, (i * std::mem::size_of::<DrawInstancedIndirectArgs>()) as u32,
)?; );
draw_with_texture(
&self.devices.device_context,
&self.pipelines.paths_pipeline,
&texture_view,
&self.context.viewport,
&self.globals.global_params_buffer,
&self.globals.sampler,
1,
)?;
} }
Ok(()) Ok(())
} }
@ -484,11 +517,13 @@ impl DirectXRenderPipelines {
std::mem::size_of::<PolychromeSprite>(), std::mem::size_of::<PolychromeSprite>(),
32, 32,
)?; )?;
let paths_indirect_draw_buffer = create_indirect_draw_buffer(device, 32)?;
Ok(Self { Ok(Self {
shadow_pipeline, shadow_pipeline,
quad_pipeline, quad_pipeline,
paths_pipeline, paths_pipeline,
paths_indirect_draw_buffer,
underline_pipeline, underline_pipeline,
mono_sprites, mono_sprites,
poly_sprites, poly_sprites,
@ -884,6 +919,27 @@ fn create_buffer_view(
Ok([view]) Ok([view])
} }
fn create_indirect_draw_buffer(device: &ID3D11Device, buffer_size: u32) -> Result<ID3D11Buffer> {
// let desc = D3D11_BUFFER_DESC {
// ByteWidth: std::mem::size_of::<DrawInstancedIndirectArgs>() as u32 * buffer_size,
// Usage: D3D11_USAGE_DYNAMIC,
// BindFlags: D3D11_BIND_INDIRECT_DRAW.0 as u32,
// MiscFlags: D3D11_RESOURCE_MISC_DRAWINDIRECT_ARGS.0 as u32,
// ..Default::default()
// };
let desc = D3D11_BUFFER_DESC {
ByteWidth: std::mem::size_of::<DrawInstancedIndirectArgs>() as u32 * buffer_size,
Usage: D3D11_USAGE_DYNAMIC,
BindFlags: D3D11_BIND_INDEX_BUFFER.0 as u32,
CPUAccessFlags: D3D11_CPU_ACCESS_WRITE.0 as u32,
MiscFlags: D3D11_RESOURCE_MISC_DRAWINDIRECT_ARGS.0 as u32,
StructureByteStride: std::mem::size_of::<DrawInstancedIndirectArgs>() as u32,
};
let mut buffer = None;
unsafe { device.CreateBuffer(&desc, None, Some(&mut buffer)) }?;
Ok(buffer.unwrap())
}
fn update_global_params( fn update_global_params(
device_context: &ID3D11DeviceContext, device_context: &ID3D11DeviceContext,
buffer: &[Option<ID3D11Buffer>; 1], buffer: &[Option<ID3D11Buffer>; 1],
@ -964,6 +1020,50 @@ fn update_buffer<T>(
Ok(()) Ok(())
} }
fn update_indirect_buffer(
device_context: &ID3D11DeviceContext,
buffer: &ID3D11Buffer,
data: &[DrawInstancedIndirectArgs],
) -> Result<()> {
unsafe {
let mut dest = std::mem::zeroed();
device_context.Map(buffer, 0, D3D11_MAP_WRITE_DISCARD, 0, Some(&mut dest))?;
std::ptr::copy_nonoverlapping(data.as_ptr(), dest.pData as _, data.len());
device_context.Unmap(buffer, 0);
}
Ok(())
}
fn prepare_indirect_draws(
device_context: &ID3D11DeviceContext,
pipeline: &PipelineState,
viewport: &[D3D11_VIEWPORT],
global_params: &[Option<ID3D11Buffer>],
topology: D3D_PRIMITIVE_TOPOLOGY,
) -> Result<()> {
unsafe {
device_context.VSSetShaderResources(1, Some(&pipeline.view));
device_context.PSSetShaderResources(1, Some(&pipeline.view));
device_context.IASetPrimitiveTopology(topology);
device_context.RSSetViewports(Some(viewport));
device_context.VSSetShader(&pipeline.vertex, None);
device_context.PSSetShader(&pipeline.fragment, None);
device_context.VSSetConstantBuffers(0, Some(global_params));
device_context.PSSetConstantBuffers(0, Some(global_params));
}
Ok(())
}
fn draw_indirect(
device_context: &ID3D11DeviceContext,
indirect_draw_buffer: &ID3D11Buffer,
offset: u32,
) {
unsafe {
device_context.DrawInstancedIndirect(indirect_draw_buffer, offset);
}
}
fn draw_normal( fn draw_normal(
device_context: &ID3D11DeviceContext, device_context: &ID3D11DeviceContext,
pipeline: &PipelineState, pipeline: &PipelineState,
@ -1026,6 +1126,7 @@ mod shader_resources {
use windows_core::{HSTRING, PCSTR}; use windows_core::{HSTRING, PCSTR};
pub(super) fn build_shader_blob(entry: &str, target: &str) -> Result<ID3DBlob> { pub(super) fn build_shader_blob(entry: &str, target: &str) -> Result<ID3DBlob> {
println!("Building shader: {}", entry);
unsafe { unsafe {
let mut entry = entry.to_owned(); let mut entry = entry.to_owned();
let mut target = target.to_owned(); let mut target = target.to_owned();
@ -1039,6 +1140,11 @@ mod shader_resources {
target.push_str("\0"); target.push_str("\0");
let entry_point = PCSTR::from_raw(entry.as_ptr()); let entry_point = PCSTR::from_raw(entry.as_ptr());
let target_cstr = PCSTR::from_raw(target.as_ptr()); let target_cstr = PCSTR::from_raw(target.as_ptr());
println!(
"Compiling shader: {} with target: {}",
entry_point.display(),
target_cstr.display()
);
#[cfg(debug_assertions)] #[cfg(debug_assertions)]
let compile_flag = D3DCOMPILE_DEBUG | D3DCOMPILE_SKIP_OPTIMIZATION; let compile_flag = D3DCOMPILE_DEBUG | D3DCOMPILE_SKIP_OPTIMIZATION;
#[cfg(not(debug_assertions))] #[cfg(not(debug_assertions))]
@ -1054,6 +1160,7 @@ mod shader_resources {
&mut compile_blob, &mut compile_blob,
Some(&mut error_blob), Some(&mut error_blob),
); );
println!("Shader compile result: {:?}", ret);
if ret.is_err() { if ret.is_err() {
let Some(error_blob) = error_blob else { let Some(error_blob) = error_blob else {
return Err(anyhow::anyhow!("{ret:?}")); return Err(anyhow::anyhow!("{ret:?}"));
@ -1064,10 +1171,9 @@ mod shader_resources {
string_len, string_len,
string_len, string_len,
); );
return Err(anyhow::anyhow!( let error_string = String::from_utf8_lossy(&error_string_encode);
"Compile error: {}", println!("Shader compile error: {}", error_string);
String::from_utf8_lossy(&error_string_encode) return Err(anyhow::anyhow!("Compile error: {}", error_string));
));
} }
Ok(compile_blob.unwrap()) Ok(compile_blob.unwrap())
} }

View file

@ -81,28 +81,36 @@ struct TransformationMatrix {
static const float M_PI_F = 3.141592653f; static const float M_PI_F = 3.141592653f;
static const float3 GRAYSCALE_FACTORS = float3(0.2126f, 0.7152f, 0.0722f); static const float3 GRAYSCALE_FACTORS = float3(0.2126f, 0.7152f, 0.0722f);
float4 to_device_position(float2 unit_vertex, Bounds bounds) { float4 to_device_position_impl(float2 position) {
float2 position = unit_vertex * bounds.size + bounds.origin;
float2 device_position = position / global_viewport_size * float2(2.0, -2.0) + float2(-1.0, 1.0); float2 device_position = position / global_viewport_size * float2(2.0, -2.0) + float2(-1.0, 1.0);
return float4(device_position, 0., 1.); return float4(device_position, 0., 1.);
} }
float4 distance_from_clip_rect(float2 unit_vertex, Bounds bounds, Bounds clip_bounds) { float4 to_device_position(float2 unit_vertex, Bounds bounds) {
float2 position = unit_vertex * bounds.size + bounds.origin; float2 position = unit_vertex * bounds.size + bounds.origin;
return to_device_position_impl(position);
}
float4 distance_from_clip_rect_impl(float2 position, Bounds clip_bounds) {
return float4(position.x - clip_bounds.origin.x, return float4(position.x - clip_bounds.origin.x,
clip_bounds.origin.x + clip_bounds.size.x - position.x, clip_bounds.origin.x + clip_bounds.size.x - position.x,
position.y - clip_bounds.origin.y, position.y - clip_bounds.origin.y,
clip_bounds.origin.y + clip_bounds.size.y - position.y); clip_bounds.origin.y + clip_bounds.size.y - position.y);
} }
float4 distance_from_clip_rect(float2 unit_vertex, Bounds bounds, Bounds clip_bounds) {
float2 position = unit_vertex * bounds.size + bounds.origin;
return distance_from_clip_rect_impl(position, clip_bounds);
}
// Convert linear RGB to sRGB // Convert linear RGB to sRGB
float3 linear_to_srgb(float3 color) { float3 linear_to_srgb(float3 color) {
return pow(color, float3(2.2)); return pow(color, float3(2.2, 2.2, 2.2));
} }
// Convert sRGB to linear RGB // Convert sRGB to linear RGB
float3 srgb_to_linear(float3 color) { float3 srgb_to_linear(float3 color) {
return pow(color, float3(1.0 / 2.2)); return pow(color, float3(1.0 / 2.2, 1.0 / 2.2, 1.0 / 2.2));
} }
/// Hsla to linear RGBA conversion. /// Hsla to linear RGBA conversion.
@ -268,23 +276,23 @@ float quad_sdf(float2 pt, Bounds bounds, Corners corner_radii) {
} }
GradientColor prepare_gradient_color(uint tag, uint color_space, Hsla solid, Hsla color0, Hsla color1) { GradientColor prepare_gradient_color(uint tag, uint color_space, Hsla solid, Hsla color0, Hsla color1) {
GradientColor res; GradientColor output;
if (tag == 0) { if (tag == 0) {
res.solid = hsla_to_rgba(solid); output.solid = hsla_to_rgba(solid);
} else if (tag == 1) { } else if (tag == 1) {
res.color0 = hsla_to_rgba(color0); output.color0 = hsla_to_rgba(color0);
res.color1 = hsla_to_rgba(color1); output.color1 = hsla_to_rgba(color1);
// Prepare color space in vertex for avoid conversion // Prepare color space in vertex for avoid conversion
// in fragment shader for performance reasons // in fragment shader for performance reasons
if (color_space == 1) { if (color_space == 1) {
// Oklab // Oklab
res.color0 = srgb_to_oklab(res.color0); output.color0 = srgb_to_oklab(output.color0);
res.color1 = srgb_to_oklab(res.color1); output.color1 = srgb_to_oklab(output.color1);
} }
} }
return res; return output;
} }
float4 gradient_color(Background background, float4 gradient_color(Background background,
@ -456,22 +464,21 @@ struct Quad {
struct QuadVertexOutput { struct QuadVertexOutput {
float4 position: SV_Position; float4 position: SV_Position;
// float4 border_color: COLOR0; nointerpolation float4 border_color: COLOR0;
float4 border_color: FLAT; nointerpolation uint quad_id: TEXCOORD0;
uint quad_id: FLAT; nointerpolation float4 background_solid: COLOR1;
float4 background_solid: FLAT; nointerpolation float4 background_color0: COLOR2;
float4 background_color0: FLAT; nointerpolation float4 background_color1: COLOR3;
float4 background_color1: FLAT;
float4 clip_distance: SV_ClipDistance; float4 clip_distance: SV_ClipDistance;
}; };
struct QuadFragmentInput { struct QuadFragmentInput {
uint quad_id: FLAT; nointerpolation uint quad_id: TEXCOORD0;
float4 position: SV_Position; float4 position: SV_Position;
float4 border_color: FLAT; nointerpolation float4 border_color: COLOR0;
float4 background_solid: FLAT; nointerpolation float4 background_solid: COLOR1;
float4 background_color0: FLAT; nointerpolation float4 background_color0: COLOR2;
float4 background_color1: FLAT; nointerpolation float4 background_color1: COLOR3;
}; };
StructuredBuffer<Quad> quads: register(t1); StructuredBuffer<Quad> quads: register(t1);
@ -566,55 +573,11 @@ float4 quad_fragment(QuadFragmentInput input): SV_Target {
return color * float4(1., 1., 1., saturate(0.5 - distance)); return color * float4(1., 1., 1., saturate(0.5 - distance));
} }
/*
**
** Path raster
**
*/
struct PathVertex { struct PathVertex {
float2 xy_position; float2 xy_position;
float2 st_position;
Bounds content_mask; Bounds content_mask;
}; };
struct PathRasterizationOutput {
float4 position: SV_Position;
float2 st_position: TEXCOORD0;
float4 clip_distances: SV_ClipDistance;
};
struct PathRasterizationInput {
float4 position: SV_Position;
float2 st_position: TEXCOORD0;
};
StructuredBuffer<PathVertex> path_vertices: register(t1);
PathRasterizationOutput path_rasterization_vertex(uint vertex_id: SV_VertexID) {
PathVertex vertex = path_vertices[vertex_id];
PathRasterizationOutput output;
float2 device_position = vertex.xy_position / global_viewport_size * float2(2.0, -2.0) + float2(-1.0, 1.0);
float2 tl = vertex.xy_position - vertex.content_mask.origin;
float2 br = vertex.content_mask.origin + vertex.content_mask.size - vertex.xy_position;
output.position = float4(device_position, 0.0, 1.0);
output.st_position = vertex.st_position;
output.clip_distances = float4(tl.x, br.x, tl.y, br.y);
return output;
}
float4 path_rasterization_fragment(PathRasterizationInput input): SV_Target {
float2 dx = ddx(input.st_position);
float2 dy = ddy(input.st_position);
float2 gradient = float2((2. * input.st_position.x) * dx.x - dx.y,
(2. * input.st_position.x) * dy.x - dy.y);
float f = (input.st_position.x * input.st_position.x) - input.st_position.y;
float distance = f / length(gradient);
float alpha = saturate(0.5 - distance);
return float4(alpha, 0., 0., 1.);
}
/* /*
** **
** Paths ** Paths
@ -624,27 +587,27 @@ float4 path_rasterization_fragment(PathRasterizationInput input): SV_Target {
struct PathSprite { struct PathSprite {
Bounds bounds; Bounds bounds;
Background color; Background color;
AtlasTile tile;
}; };
struct PathVertexOutput { struct PathVertexOutput {
float4 position: SV_Position; float4 position: SV_Position;
float2 tile_position: POSITION1; float4 clip_distance: SV_ClipDistance;
uint sprite_id: FLAT; nointerpolation uint sprite_id: TEXCOORD0;
float4 solid_color: FLAT; nointerpolation float4 solid_color: COLOR0;
float4 color0: FLAT; nointerpolation float4 color0: COLOR1;
float4 color1: FLAT; nointerpolation float4 color1: COLOR2;
}; };
StructuredBuffer<PathSprite> path_sprites: register(t1); StructuredBuffer<PathVertex> path_vertices: register(t1);
StructuredBuffer<PathSprite> path_sprites: register(t2);
PathVertexOutput paths_vertex(uint vertex_id: SV_VertexID, uint instance_id: SV_InstanceID) { PathVertexOutput paths_vertex(uint vertex_id: SV_VertexID, uint instance_id: SV_InstanceID) {
float2 unit_vertex = float2(float(vertex_id & 1u), 0.5 * float(vertex_id & 2u)); PathVertex v = path_vertices[vertex_id];
PathSprite sprite = path_sprites[instance_id]; PathSprite sprite = path_sprites[instance_id];
// Don't apply content mask because it was already accounted for when rasterizing the path.
PathVertexOutput output; PathVertexOutput output;
output.position = to_device_position(unit_vertex, sprite.bounds); output.position = to_device_position_impl(v.xy_position);
output.tile_position = to_tile_position(unit_vertex, sprite.tile); output.clip_distance = distance_from_clip_rect_impl(v.xy_position, v.content_mask);
output.sprite_id = instance_id; output.sprite_id = instance_id;
GradientColor gradient = prepare_gradient_color( GradientColor gradient = prepare_gradient_color(
@ -662,13 +625,15 @@ PathVertexOutput paths_vertex(uint vertex_id: SV_VertexID, uint instance_id: SV_
} }
float4 paths_fragment(PathVertexOutput input): SV_Target { float4 paths_fragment(PathVertexOutput input): SV_Target {
float sample = t_sprite.Sample(s_sprite, input.tile_position).r; float4 zero = 0.0;
float mask = 1.0 - abs(1.0 - sample % 2.0); if (any(input.clip_distance < zero)) {
return zero;
}
PathSprite sprite = path_sprites[input.sprite_id]; PathSprite sprite = path_sprites[input.sprite_id];
Background background = sprite.color; Background background = sprite.color;
float4 color = gradient_color(background, input.position.xy, sprite.bounds, float4 color = gradient_color(background, input.position.xy, sprite.bounds,
input.solid_color, input.color0, input.color1); input.solid_color, input.color0, input.color1);
color.a *= mask;
return color; return color;
} }