1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
//! GPU-accelerated diff bounding rect computation.
//!
//! A compute shader compares two textures (pre-stroke scratch vs post-stroke
//! canvas) and produces the tight bounding rect of all differing pixels using
//! atomic min/max. Used at stroke end to determine the exact undo region
//! without hand-tracking dab positions.
/// Initial values for the atomic bounds buffer: min = MAX, max = 0.
/// If min_x > max_x after dispatch, the textures are identical.
const BOUNDS_INIT: [u32; 4] = [u32::MAX, u32::MAX, 0, 0];
pub struct DiffRectPass {
pipeline: wgpu::ComputePipeline,
bind_group_layout: wgpu::BindGroupLayout,
pending: Option<PendingDiff>,
}
struct PendingDiff {
staging: wgpu::Buffer,
rx: Option<std::sync::mpsc::Receiver<Result<(), wgpu::BufferAsyncError>>>,
/// Canvas-space extent of the layer at the time of `request`. Used to
/// translate the shader's layer-local bounding rect back to canvas
/// coords on `poll` — must be captured at request time so the result
/// remains correct even if the layer grows between request and poll.
layer_canvas_extent: crate::coord::CanvasRect,
}
/// Uniform buffer layout matching the shader's `Params` struct.
#[repr(C)]
#[derive(Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
struct Params {
width: u32,
height: u32,
}
impl DiffRectPass {
pub fn new(device: &wgpu::Device) -> Self {
let shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {
label: Some("diff-rect-shader"),
source: wgpu::ShaderSource::Wgsl(include_str!("../../shaders/diff_rect.wgsl").into()),
});
let bind_group_layout = device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
label: Some("diff-rect-bgl"),
entries: &[
// binding 0: texture A (scratch / pre-stroke)
wgpu::BindGroupLayoutEntry {
binding: 0,
visibility: wgpu::ShaderStages::COMPUTE,
ty: wgpu::BindingType::Texture {
sample_type: wgpu::TextureSampleType::Float { filterable: false },
view_dimension: wgpu::TextureViewDimension::D2,
multisampled: false,
},
count: None,
},
// binding 1: texture B (current canvas)
wgpu::BindGroupLayoutEntry {
binding: 1,
visibility: wgpu::ShaderStages::COMPUTE,
ty: wgpu::BindingType::Texture {
sample_type: wgpu::TextureSampleType::Float { filterable: false },
view_dimension: wgpu::TextureViewDimension::D2,
multisampled: false,
},
count: None,
},
// binding 2: atomic bounds storage buffer
wgpu::BindGroupLayoutEntry {
binding: 2,
visibility: wgpu::ShaderStages::COMPUTE,
ty: wgpu::BindingType::Buffer {
ty: wgpu::BufferBindingType::Storage { read_only: false },
has_dynamic_offset: false,
min_binding_size: None,
},
count: None,
},
// binding 3: params uniform
wgpu::BindGroupLayoutEntry {
binding: 3,
visibility: wgpu::ShaderStages::COMPUTE,
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("diff-rect-pipeline-layout"),
bind_group_layouts: &[Some(&bind_group_layout)],
immediate_size: 0,
});
let pipeline = device.create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
label: Some("diff-rect-pipeline"),
layout: Some(&pipeline_layout),
module: &shader,
entry_point: Some("main"),
compilation_options: Default::default(),
cache: None,
});
DiffRectPass {
pipeline,
bind_group_layout,
pending: None,
}
}
/// Dispatch the diff compute shader comparing two textures.
///
/// `scratch_view` is the pre-stroke snapshot, `current_view` is the
/// post-stroke canvas. `layer_canvas_extent` is the canvas-space rect
/// occupied by the layer at request time — used to translate the
/// layer-local result back to canvas coords when [`poll`] resolves.
/// Capturing it at request time keeps the result correct even if the
/// layer grows before the result is read. Results arrive asynchronously
/// via [`poll`].
pub fn request(
&mut self,
device: &wgpu::Device,
queue: &wgpu::Queue,
scratch_view: &wgpu::TextureView,
current_view: &wgpu::TextureView,
layer_canvas_extent: crate::coord::CanvasRect,
) {
let width = layer_canvas_extent.width;
let height = layer_canvas_extent.height;
// Storage buffer for atomic results (16 bytes).
let storage_buf = device.create_buffer(&wgpu::BufferDescriptor {
label: Some("diff-rect-storage"),
size: 16,
usage: wgpu::BufferUsages::STORAGE | wgpu::BufferUsages::COPY_SRC,
mapped_at_creation: true,
});
{
let mut mapping = storage_buf.slice(..).get_mapped_range_mut();
mapping.copy_from_slice(bytemuck::bytes_of(&BOUNDS_INIT));
}
storage_buf.unmap();
// Staging buffer for CPU readback (16 bytes).
let staging_buf = device.create_buffer(&wgpu::BufferDescriptor {
label: Some("diff-rect-staging"),
size: 16,
usage: wgpu::BufferUsages::MAP_READ | wgpu::BufferUsages::COPY_DST,
mapped_at_creation: false,
});
// Params uniform.
let params = Params { width, height };
let param_buf = device.create_buffer(&wgpu::BufferDescriptor {
label: Some("diff-rect-params"),
size: std::mem::size_of::<Params>() as u64,
usage: wgpu::BufferUsages::UNIFORM,
mapped_at_creation: true,
});
{
let mut mapping = param_buf.slice(..).get_mapped_range_mut();
mapping.copy_from_slice(bytemuck::bytes_of(¶ms));
}
param_buf.unmap();
let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
label: Some("diff-rect-bg"),
layout: &self.bind_group_layout,
entries: &[
wgpu::BindGroupEntry {
binding: 0,
resource: wgpu::BindingResource::TextureView(scratch_view),
},
wgpu::BindGroupEntry {
binding: 1,
resource: wgpu::BindingResource::TextureView(current_view),
},
wgpu::BindGroupEntry {
binding: 2,
resource: storage_buf.as_entire_binding(),
},
wgpu::BindGroupEntry {
binding: 3,
resource: param_buf.as_entire_binding(),
},
],
});
let mut encoder = device.create_command_encoder(&wgpu::CommandEncoderDescriptor {
label: Some("diff-rect-compute"),
});
{
let mut pass = encoder.begin_compute_pass(&wgpu::ComputePassDescriptor {
label: Some("diff-rect"),
timestamp_writes: None,
});
pass.set_pipeline(&self.pipeline);
pass.set_bind_group(0, Some(&bind_group), &[]);
let wg_x = width.div_ceil(16);
let wg_y = height.div_ceil(16);
pass.dispatch_workgroups(wg_x, wg_y, 1);
}
// Copy storage -> staging for CPU readback.
encoder.copy_buffer_to_buffer(&storage_buf, 0, &staging_buf, 0, 16);
queue.submit([encoder.finish()]);
self.pending = Some(PendingDiff {
staging: staging_buf,
rx: None,
layer_canvas_extent,
});
}
/// True if a diff result is pending.
pub fn is_pending(&self) -> bool {
self.pending.is_some()
}
/// Poll for the diff result. Returns `Some(Some(rect))` when ready,
/// `Some(None)` if the textures are identical, or `None` if still pending.
///
/// The rect is returned in canvas coords: the shader produces
/// layer-local bounds, which are translated using the layer's canvas
/// extent captured at [`request`](Self::request) time. The result
/// therefore remains valid through layer growth between request and poll.
pub fn poll(&mut self, device: &wgpu::Device) -> Option<Option<crate::coord::CanvasRect>> {
let pending = self.pending.as_mut()?;
// Begin mapping if not started.
if pending.rx.is_none() {
let slice = pending.staging.slice(..);
let (tx, rx) = std::sync::mpsc::sync_channel(1);
slice.map_async(wgpu::MapMode::Read, move |result| {
let _ = tx.send(result);
});
pending.rx = Some(rx);
}
let _ = device.poll(wgpu::PollType::Poll);
let ready = pending.rx.as_ref().unwrap().try_recv().ok();
match ready {
Some(Ok(())) => {
let p = self.pending.take().unwrap();
let slice = p.staging.slice(..);
let mapped = slice.get_mapped_range();
let raw: [u32; 4] = *bytemuck::from_bytes(&mapped[..16]);
drop(mapped);
p.staging.unmap();
let [min_x, min_y, max_x, max_y] = raw;
if min_x <= max_x && min_y <= max_y {
// +1 because max is inclusive pixel coordinate.
// Translate layer-local bounds to canvas coords using the
// extent captured at request time.
let canvas_x = p.layer_canvas_extent.origin.x + min_x as i32;
let canvas_y = p.layer_canvas_extent.origin.y + min_y as i32;
Some(Some(crate::coord::CanvasRect::from_xywh(
canvas_x,
canvas_y,
max_x - min_x + 1,
max_y - min_y + 1,
)))
} else {
// Textures are identical — no diff.
Some(None)
}
}
Some(Err(e)) => {
log::error!("diff rect buffer mapping failed: {e}");
self.pending = None;
Some(None)
}
None => None,
}
}
}