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
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
#![allow(unsafe_code)]

use std::{borrow::Cow, num::NonZeroU64, ops::Range};

use ahash::HashMap;
use epaint::{emath::NumExt, PaintCallbackInfo, Primitive, Vertex};

use wgpu::util::DeviceExt as _;

/// You can use this for storage when implementing [`CallbackTrait`].
pub type CallbackResources = type_map::concurrent::TypeMap;

/// You can use this to do custom [`wgpu`] rendering in an egui app.
///
/// Implement [`CallbackTrait`] and call [`Callback::new_paint_callback`].
///
/// This can be turned into a [`epaint::PaintCallback`] and [`epaint::Shape`].
pub struct Callback(Box<dyn CallbackTrait>);

impl Callback {
    /// Creates a new [`epaint::PaintCallback`] from a callback trait instance.
    pub fn new_paint_callback(
        rect: epaint::emath::Rect,
        callback: impl CallbackTrait + 'static,
    ) -> epaint::PaintCallback {
        epaint::PaintCallback {
            rect,
            callback: std::sync::Arc::new(Self(Box::new(callback))),
        }
    }
}

/// A callback trait that can be used to compose an [`epaint::PaintCallback`] via [`Callback`]
/// for custom WGPU rendering.
///
/// Callbacks in [`Renderer`] are done in three steps:
/// * [`CallbackTrait::prepare`]: called for all registered callbacks before the main egui render pass.
/// * [`CallbackTrait::finish_prepare`]: called for all registered callbacks after all callbacks finished calling prepare.
/// * [`CallbackTrait::paint`]: called for all registered callbacks during the main egui render pass.
///
/// Each callback has access to an instance of [`CallbackResources`] that is stored in the [`Renderer`].
/// This can be used to store wgpu resources that need to be accessed during the [`CallbackTrait::paint`] step.
///
/// The callbacks implementing [`CallbackTrait`] itself must always be Send + Sync, but resources stored in
/// [`Renderer::callback_resources`] are not required to implement Send + Sync when building for wasm.
/// (this is because wgpu stores references to the JS heap in most of its resources which can not be shared with other threads).
///
///
/// # Command submission
///
/// ## Command Encoder
///
/// The passed-in [`wgpu::CommandEncoder`] is egui's and can be used directly to register
/// wgpu commands for simple use cases.
/// This allows reusing the same [`wgpu::CommandEncoder`] for all callbacks and egui
/// rendering itself.
///
/// ## Command Buffers
///
/// For more complicated use cases, one can also return a list of arbitrary
/// [`wgpu::CommandBuffer`]s and have complete control over how they get created and fed.
/// In particular, this gives an opportunity to parallelize command registration and
/// prevents a faulty callback from poisoning the main wgpu pipeline.
///
/// When using eframe, the main egui command buffer, as well as all user-defined
/// command buffers returned by this function, are guaranteed to all be submitted
/// at once in a single call.
///
/// Command Buffers returned by [`CallbackTrait::finish_prepare`] will always be issued *after*
/// those returned by [`CallbackTrait::prepare`].
/// Order within command buffers returned by [`CallbackTrait::prepare`] is dependent
/// on the order the respective [`epaint::Shape::Callback`]s were submitted in.
///
/// # Example
///
/// See the [`custom3d_wgpu`](https://github.com/emilk/egui/blob/master/crates/egui_demo_app/src/apps/custom3d_wgpu.rs) demo source for a detailed usage example.
pub trait CallbackTrait: Send + Sync {
    fn prepare(
        &self,
        _device: &wgpu::Device,
        _queue: &wgpu::Queue,
        _screen_descriptor: &ScreenDescriptor,
        _egui_encoder: &mut wgpu::CommandEncoder,
        _callback_resources: &mut CallbackResources,
    ) -> Vec<wgpu::CommandBuffer> {
        Vec::new()
    }

    /// Called after all [`CallbackTrait::prepare`] calls are done.
    fn finish_prepare(
        &self,
        _device: &wgpu::Device,
        _queue: &wgpu::Queue,
        _egui_encoder: &mut wgpu::CommandEncoder,
        _callback_resources: &mut CallbackResources,
    ) -> Vec<wgpu::CommandBuffer> {
        Vec::new()
    }

    /// Called after all [`CallbackTrait::finish_prepare`] calls are done.
    ///
    /// It is given access to the [`wgpu::RenderPass`] so that it can issue draw commands
    /// into the same [`wgpu::RenderPass`] that is used for all other egui elements.
    fn paint<'a>(
        &'a self,
        info: PaintCallbackInfo,
        render_pass: &mut wgpu::RenderPass<'a>,
        callback_resources: &'a CallbackResources,
    );
}

/// Information about the screen used for rendering.
pub struct ScreenDescriptor {
    /// Size of the window in physical pixels.
    pub size_in_pixels: [u32; 2],

    /// HiDPI scale factor (pixels per point).
    pub pixels_per_point: f32,
}

impl ScreenDescriptor {
    /// size in "logical" points
    fn screen_size_in_points(&self) -> [f32; 2] {
        [
            self.size_in_pixels[0] as f32 / self.pixels_per_point,
            self.size_in_pixels[1] as f32 / self.pixels_per_point,
        ]
    }
}

/// Uniform buffer used when rendering.
#[derive(Clone, Copy, Debug, bytemuck::Pod, bytemuck::Zeroable)]
#[repr(C)]
struct UniformBuffer {
    screen_size_in_points: [f32; 2],
    // Uniform buffers need to be at least 16 bytes in WebGL.
    // See https://github.com/gfx-rs/wgpu/issues/2072
    _padding: [u32; 2],
}

impl PartialEq for UniformBuffer {
    fn eq(&self, other: &Self) -> bool {
        self.screen_size_in_points == other.screen_size_in_points
    }
}

struct SlicedBuffer {
    buffer: wgpu::Buffer,
    slices: Vec<Range<usize>>,
    capacity: wgpu::BufferAddress,
}

/// Renderer for a egui based GUI.
pub struct Renderer {
    pipeline: wgpu::RenderPipeline,

    index_buffer: SlicedBuffer,
    vertex_buffer: SlicedBuffer,

    uniform_buffer: wgpu::Buffer,
    previous_uniform_buffer_content: UniformBuffer,
    uniform_bind_group: wgpu::BindGroup,
    texture_bind_group_layout: wgpu::BindGroupLayout,

    /// Map of egui texture IDs to textures and their associated bindgroups (texture view +
    /// sampler). The texture may be None if the `TextureId` is just a handle to a user-provided
    /// sampler.
    textures: HashMap<epaint::TextureId, (Option<wgpu::Texture>, wgpu::BindGroup)>,
    next_user_texture_id: u64,
    samplers: HashMap<epaint::textures::TextureOptions, wgpu::Sampler>,

    /// Storage for resources shared with all invocations of [`CallbackTrait`]'s methods.
    ///
    /// See also [`CallbackTrait`].
    pub callback_resources: CallbackResources,
}

impl Renderer {
    /// Creates a renderer for a egui UI.
    ///
    /// `output_color_format` should preferably be [`wgpu::TextureFormat::Rgba8Unorm`] or
    /// [`wgpu::TextureFormat::Bgra8Unorm`], i.e. in gamma-space.
    pub fn new(
        device: &wgpu::Device,
        output_color_format: wgpu::TextureFormat,
        output_depth_format: Option<wgpu::TextureFormat>,
        msaa_samples: u32,
    ) -> Self {
        crate::profile_function!();

        let shader = wgpu::ShaderModuleDescriptor {
            label: Some("egui"),
            source: wgpu::ShaderSource::Wgsl(Cow::Borrowed(include_str!("egui.wgsl"))),
        };
        let module = {
            crate::profile_scope!("create_shader_module");
            device.create_shader_module(shader)
        };

        let uniform_buffer = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
            label: Some("egui_uniform_buffer"),
            contents: bytemuck::cast_slice(&[UniformBuffer {
                screen_size_in_points: [0.0, 0.0],
                _padding: Default::default(),
            }]),
            usage: wgpu::BufferUsages::UNIFORM | wgpu::BufferUsages::COPY_DST,
        });

        let uniform_bind_group_layout = {
            crate::profile_scope!("create_bind_group_layout");
            device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
                label: Some("egui_uniform_bind_group_layout"),
                entries: &[wgpu::BindGroupLayoutEntry {
                    binding: 0,
                    visibility: wgpu::ShaderStages::VERTEX,
                    ty: wgpu::BindingType::Buffer {
                        has_dynamic_offset: false,
                        min_binding_size: NonZeroU64::new(std::mem::size_of::<UniformBuffer>() as _),
                        ty: wgpu::BufferBindingType::Uniform,
                    },
                    count: None,
                }],
            })
        };

        let uniform_bind_group = {
            crate::profile_scope!("create_bind_group");
            device.create_bind_group(&wgpu::BindGroupDescriptor {
                label: Some("egui_uniform_bind_group"),
                layout: &uniform_bind_group_layout,
                entries: &[wgpu::BindGroupEntry {
                    binding: 0,
                    resource: wgpu::BindingResource::Buffer(wgpu::BufferBinding {
                        buffer: &uniform_buffer,
                        offset: 0,
                        size: None,
                    }),
                }],
            })
        };

        let texture_bind_group_layout = {
            crate::profile_scope!("create_bind_group_layout");
            device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
                label: Some("egui_texture_bind_group_layout"),
                entries: &[
                    wgpu::BindGroupLayoutEntry {
                        binding: 0,
                        visibility: wgpu::ShaderStages::FRAGMENT,
                        ty: wgpu::BindingType::Texture {
                            multisampled: false,
                            sample_type: wgpu::TextureSampleType::Float { filterable: true },
                            view_dimension: wgpu::TextureViewDimension::D2,
                        },
                        count: None,
                    },
                    wgpu::BindGroupLayoutEntry {
                        binding: 1,
                        visibility: wgpu::ShaderStages::FRAGMENT,
                        ty: wgpu::BindingType::Sampler(wgpu::SamplerBindingType::Filtering),
                        count: None,
                    },
                ],
            })
        };

        let pipeline_layout = device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
            label: Some("egui_pipeline_layout"),
            bind_group_layouts: &[&uniform_bind_group_layout, &texture_bind_group_layout],
            push_constant_ranges: &[],
        });

        let depth_stencil = output_depth_format.map(|format| wgpu::DepthStencilState {
            format,
            depth_write_enabled: false,
            depth_compare: wgpu::CompareFunction::Always,
            stencil: wgpu::StencilState::default(),
            bias: wgpu::DepthBiasState::default(),
        });

        let pipeline = {
            crate::profile_scope!("create_render_pipeline");
            device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
                label: Some("egui_pipeline"),
                layout: Some(&pipeline_layout),
                vertex: wgpu::VertexState {
                    entry_point: "vs_main",
                    module: &module,
                    buffers: &[wgpu::VertexBufferLayout {
                        array_stride: 5 * 4,
                        step_mode: wgpu::VertexStepMode::Vertex,
                        // 0: vec2 position
                        // 1: vec2 texture coordinates
                        // 2: uint color
                        attributes: &wgpu::vertex_attr_array![0 => Float32x2, 1 => Float32x2, 2 => Uint32],
                    }],
                    compilation_options: wgpu::PipelineCompilationOptions::default()
                },
                primitive: wgpu::PrimitiveState {
                    topology: wgpu::PrimitiveTopology::TriangleList,
                    unclipped_depth: false,
                    conservative: false,
                    cull_mode: None,
                    front_face: wgpu::FrontFace::default(),
                    polygon_mode: wgpu::PolygonMode::default(),
                    strip_index_format: None,
                },
                depth_stencil,
                multisample: wgpu::MultisampleState {
                    alpha_to_coverage_enabled: false,
                    count: msaa_samples,
                    mask: !0,
                },

                fragment: Some(wgpu::FragmentState {
                    module: &module,
                    entry_point: if output_color_format.is_srgb() {
                        log::warn!("Detected a linear (sRGBA aware) framebuffer {:?}. egui prefers Rgba8Unorm or Bgra8Unorm", output_color_format);
                        "fs_main_linear_framebuffer"
                    } else {
                        "fs_main_gamma_framebuffer" // this is what we prefer
                    },
                    targets: &[Some(wgpu::ColorTargetState {
                        format: output_color_format,
                        blend: Some(wgpu::BlendState {
                            color: wgpu::BlendComponent {
                                src_factor: wgpu::BlendFactor::One,
                                dst_factor: wgpu::BlendFactor::OneMinusSrcAlpha,
                                operation: wgpu::BlendOperation::Add,
                            },
                            alpha: wgpu::BlendComponent {
                                src_factor: wgpu::BlendFactor::OneMinusDstAlpha,
                                dst_factor: wgpu::BlendFactor::One,
                                operation: wgpu::BlendOperation::Add,
                            },
                        }),
                        write_mask: wgpu::ColorWrites::ALL,
                    })],
                    compilation_options: wgpu::PipelineCompilationOptions::default()
                }),
                multiview: None,
            }
        )
        };

        const VERTEX_BUFFER_START_CAPACITY: wgpu::BufferAddress =
            (std::mem::size_of::<Vertex>() * 1024) as _;
        const INDEX_BUFFER_START_CAPACITY: wgpu::BufferAddress =
            (std::mem::size_of::<u32>() * 1024 * 3) as _;

        Self {
            pipeline,
            vertex_buffer: SlicedBuffer {
                buffer: create_vertex_buffer(device, VERTEX_BUFFER_START_CAPACITY),
                slices: Vec::with_capacity(64),
                capacity: VERTEX_BUFFER_START_CAPACITY,
            },
            index_buffer: SlicedBuffer {
                buffer: create_index_buffer(device, INDEX_BUFFER_START_CAPACITY),
                slices: Vec::with_capacity(64),
                capacity: INDEX_BUFFER_START_CAPACITY,
            },
            uniform_buffer,
            // Buffers on wgpu are zero initialized, so this is indeed its current state!
            previous_uniform_buffer_content: UniformBuffer {
                screen_size_in_points: [0.0, 0.0],
                _padding: [0, 0],
            },
            uniform_bind_group,
            texture_bind_group_layout,
            textures: HashMap::default(),
            next_user_texture_id: 0,
            samplers: HashMap::default(),
            callback_resources: CallbackResources::default(),
        }
    }

    /// Executes the egui renderer onto an existing wgpu renderpass.
    pub fn render<'rp>(
        &'rp self,
        render_pass: &mut wgpu::RenderPass<'rp>,
        paint_jobs: &'rp [epaint::ClippedPrimitive],
        screen_descriptor: &ScreenDescriptor,
    ) {
        crate::profile_function!();

        let pixels_per_point = screen_descriptor.pixels_per_point;
        let size_in_pixels = screen_descriptor.size_in_pixels;

        // Whether or not we need to reset the render pass because a paint callback has just
        // run.
        let mut needs_reset = true;

        let mut index_buffer_slices = self.index_buffer.slices.iter();
        let mut vertex_buffer_slices = self.vertex_buffer.slices.iter();

        for epaint::ClippedPrimitive {
            clip_rect,
            primitive,
        } in paint_jobs
        {
            if needs_reset {
                render_pass.set_viewport(
                    0.0,
                    0.0,
                    size_in_pixels[0] as f32,
                    size_in_pixels[1] as f32,
                    0.0,
                    1.0,
                );
                render_pass.set_pipeline(&self.pipeline);
                render_pass.set_bind_group(0, &self.uniform_bind_group, &[]);
                needs_reset = false;
            }

            {
                let rect = ScissorRect::new(clip_rect, pixels_per_point, size_in_pixels);

                if rect.width == 0 || rect.height == 0 {
                    // Skip rendering zero-sized clip areas.
                    if let Primitive::Mesh(_) = primitive {
                        // If this is a mesh, we need to advance the index and vertex buffer iterators:
                        index_buffer_slices.next().unwrap();
                        vertex_buffer_slices.next().unwrap();
                    }
                    continue;
                }

                render_pass.set_scissor_rect(rect.x, rect.y, rect.width, rect.height);
            }

            match primitive {
                Primitive::Mesh(mesh) => {
                    let index_buffer_slice = index_buffer_slices.next().unwrap();
                    let vertex_buffer_slice = vertex_buffer_slices.next().unwrap();

                    if let Some((_texture, bind_group)) = self.textures.get(&mesh.texture_id) {
                        render_pass.set_bind_group(1, bind_group, &[]);
                        render_pass.set_index_buffer(
                            self.index_buffer.buffer.slice(
                                index_buffer_slice.start as u64..index_buffer_slice.end as u64,
                            ),
                            wgpu::IndexFormat::Uint32,
                        );
                        render_pass.set_vertex_buffer(
                            0,
                            self.vertex_buffer.buffer.slice(
                                vertex_buffer_slice.start as u64..vertex_buffer_slice.end as u64,
                            ),
                        );
                        render_pass.draw_indexed(0..mesh.indices.len() as u32, 0, 0..1);
                    } else {
                        log::warn!("Missing texture: {:?}", mesh.texture_id);
                    }
                }
                Primitive::Callback(callback) => {
                    let Some(cbfn) = callback.callback.downcast_ref::<Callback>() else {
                        // We already warned in the `prepare` callback
                        continue;
                    };

                    let info = PaintCallbackInfo {
                        viewport: callback.rect,
                        clip_rect: *clip_rect,
                        pixels_per_point,
                        screen_size_px: size_in_pixels,
                    };

                    let viewport_px = info.viewport_in_pixels();
                    if viewport_px.width_px > 0 && viewport_px.height_px > 0 {
                        crate::profile_scope!("callback");

                        needs_reset = true;

                        // We're setting a default viewport for the render pass as a
                        // courtesy for the user, so that they don't have to think about
                        // it in the simple case where they just want to fill the whole
                        // paint area.
                        //
                        // The user still has the possibility of setting their own custom
                        // viewport during the paint callback, effectively overriding this
                        // one.
                        render_pass.set_viewport(
                            viewport_px.left_px as f32,
                            viewport_px.top_px as f32,
                            viewport_px.width_px as f32,
                            viewport_px.height_px as f32,
                            0.0,
                            1.0,
                        );

                        cbfn.0.paint(info, render_pass, &self.callback_resources);
                    }
                }
            }
        }

        render_pass.set_scissor_rect(0, 0, size_in_pixels[0], size_in_pixels[1]);
    }

    /// Should be called before [`Self::render`].
    pub fn update_texture(
        &mut self,
        device: &wgpu::Device,
        queue: &wgpu::Queue,
        id: epaint::TextureId,
        image_delta: &epaint::ImageDelta,
    ) {
        crate::profile_function!();

        let width = image_delta.image.width() as u32;
        let height = image_delta.image.height() as u32;

        let size = wgpu::Extent3d {
            width,
            height,
            depth_or_array_layers: 1,
        };

        let data_color32 = match &image_delta.image {
            epaint::ImageData::Color(image) => {
                assert_eq!(
                    width as usize * height as usize,
                    image.pixels.len(),
                    "Mismatch between texture size and texel count"
                );
                Cow::Borrowed(&image.pixels)
            }
            epaint::ImageData::Font(image) => {
                assert_eq!(
                    width as usize * height as usize,
                    image.pixels.len(),
                    "Mismatch between texture size and texel count"
                );
                crate::profile_scope!("font -> sRGBA");
                Cow::Owned(image.srgba_pixels(None).collect::<Vec<egui::Color32>>())
            }
        };
        let data_bytes: &[u8] = bytemuck::cast_slice(data_color32.as_slice());

        let queue_write_data_to_texture = |texture, origin| {
            crate::profile_scope!("write_texture");
            queue.write_texture(
                wgpu::ImageCopyTexture {
                    texture,
                    mip_level: 0,
                    origin,
                    aspect: wgpu::TextureAspect::All,
                },
                data_bytes,
                wgpu::ImageDataLayout {
                    offset: 0,
                    bytes_per_row: Some(4 * width),
                    rows_per_image: Some(height),
                },
                size,
            );
        };

        if let Some(pos) = image_delta.pos {
            // update the existing texture
            let (texture, _bind_group) = self
                .textures
                .get(&id)
                .expect("Tried to update a texture that has not been allocated yet.");
            let origin = wgpu::Origin3d {
                x: pos[0] as u32,
                y: pos[1] as u32,
                z: 0,
            };
            queue_write_data_to_texture(
                texture.as_ref().expect("Tried to update user texture."),
                origin,
            );
        } else {
            // allocate a new texture
            // Use same label for all resources associated with this texture id (no point in retyping the type)
            let label_str = format!("egui_texid_{id:?}");
            let label = Some(label_str.as_str());
            let texture = {
                crate::profile_scope!("create_texture");
                device.create_texture(&wgpu::TextureDescriptor {
                    label,
                    size,
                    mip_level_count: 1,
                    sample_count: 1,
                    dimension: wgpu::TextureDimension::D2,
                    format: wgpu::TextureFormat::Rgba8UnormSrgb, // Minspec for wgpu WebGL emulation is WebGL2, so this should always be supported.
                    usage: wgpu::TextureUsages::TEXTURE_BINDING | wgpu::TextureUsages::COPY_DST,
                    view_formats: &[wgpu::TextureFormat::Rgba8UnormSrgb],
                })
            };
            let sampler = self
                .samplers
                .entry(image_delta.options)
                .or_insert_with(|| create_sampler(image_delta.options, device));
            let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
                label,
                layout: &self.texture_bind_group_layout,
                entries: &[
                    wgpu::BindGroupEntry {
                        binding: 0,
                        resource: wgpu::BindingResource::TextureView(
                            &texture.create_view(&wgpu::TextureViewDescriptor::default()),
                        ),
                    },
                    wgpu::BindGroupEntry {
                        binding: 1,
                        resource: wgpu::BindingResource::Sampler(sampler),
                    },
                ],
            });
            let origin = wgpu::Origin3d::ZERO;
            queue_write_data_to_texture(&texture, origin);
            self.textures.insert(id, (Some(texture), bind_group));
        };
    }

    pub fn free_texture(&mut self, id: &epaint::TextureId) {
        self.textures.remove(id);
    }

    /// Get the WGPU texture and bind group associated to a texture that has been allocated by egui.
    ///
    /// This could be used by custom paint hooks to render images that have been added through
    /// [`epaint::Context::load_texture`](https://docs.rs/egui/latest/egui/struct.Context.html#method.load_texture).
    pub fn texture(
        &self,
        id: &epaint::TextureId,
    ) -> Option<&(Option<wgpu::Texture>, wgpu::BindGroup)> {
        self.textures.get(id)
    }

    /// Registers a [`wgpu::Texture`] with a [`epaint::TextureId`].
    ///
    /// This enables the application to reference the texture inside an image ui element.
    /// This effectively enables off-screen rendering inside the egui UI. Texture must have
    /// the texture format [`wgpu::TextureFormat::Rgba8UnormSrgb`].
    pub fn register_native_texture(
        &mut self,
        device: &wgpu::Device,
        texture: &wgpu::TextureView,
        texture_filter: wgpu::FilterMode,
    ) -> epaint::TextureId {
        self.register_native_texture_with_sampler_options(
            device,
            texture,
            wgpu::SamplerDescriptor {
                label: Some(format!("egui_user_image_{}", self.next_user_texture_id).as_str()),
                mag_filter: texture_filter,
                min_filter: texture_filter,
                ..Default::default()
            },
        )
    }

    /// Registers a [`wgpu::Texture`] with an existing [`epaint::TextureId`].
    ///
    /// This enables applications to reuse [`epaint::TextureId`]s.
    pub fn update_egui_texture_from_wgpu_texture(
        &mut self,
        device: &wgpu::Device,
        texture: &wgpu::TextureView,
        texture_filter: wgpu::FilterMode,
        id: epaint::TextureId,
    ) {
        self.update_egui_texture_from_wgpu_texture_with_sampler_options(
            device,
            texture,
            wgpu::SamplerDescriptor {
                label: Some(format!("egui_user_image_{}", self.next_user_texture_id).as_str()),
                mag_filter: texture_filter,
                min_filter: texture_filter,
                ..Default::default()
            },
            id,
        );
    }

    /// Registers a [`wgpu::Texture`] with a [`epaint::TextureId`] while also accepting custom
    /// [`wgpu::SamplerDescriptor`] options.
    ///
    /// This allows applications to specify individual minification/magnification filters as well as
    /// custom mipmap and tiling options.
    ///
    /// The texture must have the format [`wgpu::TextureFormat::Rgba8UnormSrgb`].
    /// Any compare function supplied in the [`wgpu::SamplerDescriptor`] will be ignored.
    #[allow(clippy::needless_pass_by_value)] // false positive
    pub fn register_native_texture_with_sampler_options(
        &mut self,
        device: &wgpu::Device,
        texture: &wgpu::TextureView,
        sampler_descriptor: wgpu::SamplerDescriptor<'_>,
    ) -> epaint::TextureId {
        crate::profile_function!();

        let sampler = device.create_sampler(&wgpu::SamplerDescriptor {
            compare: None,
            ..sampler_descriptor
        });

        let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
            label: Some(format!("egui_user_image_{}", self.next_user_texture_id).as_str()),
            layout: &self.texture_bind_group_layout,
            entries: &[
                wgpu::BindGroupEntry {
                    binding: 0,
                    resource: wgpu::BindingResource::TextureView(texture),
                },
                wgpu::BindGroupEntry {
                    binding: 1,
                    resource: wgpu::BindingResource::Sampler(&sampler),
                },
            ],
        });

        let id = epaint::TextureId::User(self.next_user_texture_id);
        self.textures.insert(id, (None, bind_group));
        self.next_user_texture_id += 1;

        id
    }

    /// Registers a [`wgpu::Texture`] with an existing [`epaint::TextureId`] while also accepting custom
    /// [`wgpu::SamplerDescriptor`] options.
    ///
    /// This allows applications to reuse [`epaint::TextureId`]s created with custom sampler options.
    #[allow(clippy::needless_pass_by_value)] // false positive
    pub fn update_egui_texture_from_wgpu_texture_with_sampler_options(
        &mut self,
        device: &wgpu::Device,
        texture: &wgpu::TextureView,
        sampler_descriptor: wgpu::SamplerDescriptor<'_>,
        id: epaint::TextureId,
    ) {
        crate::profile_function!();

        let (_user_texture, user_texture_binding) = self
            .textures
            .get_mut(&id)
            .expect("Tried to update a texture that has not been allocated yet.");

        let sampler = device.create_sampler(&wgpu::SamplerDescriptor {
            compare: None,
            ..sampler_descriptor
        });

        let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
            label: Some(format!("egui_user_image_{}", self.next_user_texture_id).as_str()),
            layout: &self.texture_bind_group_layout,
            entries: &[
                wgpu::BindGroupEntry {
                    binding: 0,
                    resource: wgpu::BindingResource::TextureView(texture),
                },
                wgpu::BindGroupEntry {
                    binding: 1,
                    resource: wgpu::BindingResource::Sampler(&sampler),
                },
            ],
        });

        *user_texture_binding = bind_group;
    }

    /// Uploads the uniform, vertex and index data used by the renderer.
    /// Should be called before [`Self::render`].
    ///
    /// Returns all user-defined command buffers gathered from [`CallbackTrait::prepare`] & [`CallbackTrait::finish_prepare`] callbacks.
    pub fn update_buffers(
        &mut self,
        device: &wgpu::Device,
        queue: &wgpu::Queue,
        encoder: &mut wgpu::CommandEncoder,
        paint_jobs: &[epaint::ClippedPrimitive],
        screen_descriptor: &ScreenDescriptor,
    ) -> Vec<wgpu::CommandBuffer> {
        crate::profile_function!();

        let screen_size_in_points = screen_descriptor.screen_size_in_points();

        let uniform_buffer_content = UniformBuffer {
            screen_size_in_points,
            _padding: Default::default(),
        };
        if uniform_buffer_content != self.previous_uniform_buffer_content {
            crate::profile_scope!("update uniforms");
            queue.write_buffer(
                &self.uniform_buffer,
                0,
                bytemuck::cast_slice(&[uniform_buffer_content]),
            );
            self.previous_uniform_buffer_content = uniform_buffer_content;
        }

        // Determine how many vertices & indices need to be rendered, and gather prepare callbacks
        let mut callbacks = Vec::new();
        let (vertex_count, index_count) = {
            crate::profile_scope!("count_vertices_indices");
            paint_jobs.iter().fold((0, 0), |acc, clipped_primitive| {
                match &clipped_primitive.primitive {
                    Primitive::Mesh(mesh) => {
                        (acc.0 + mesh.vertices.len(), acc.1 + mesh.indices.len())
                    }
                    Primitive::Callback(callback) => {
                        if let Some(c) = callback.callback.downcast_ref::<Callback>() {
                            callbacks.push(c.0.as_ref());
                        } else {
                            log::warn!("Unknown paint callback: expected `egui_wgpu::Callback`");
                        };
                        acc
                    }
                }
            })
        };

        if index_count > 0 {
            crate::profile_scope!("indices", index_count.to_string());

            self.index_buffer.slices.clear();

            let required_index_buffer_size = (std::mem::size_of::<u32>() * index_count) as u64;
            if self.index_buffer.capacity < required_index_buffer_size {
                // Resize index buffer if needed.
                self.index_buffer.capacity =
                    (self.index_buffer.capacity * 2).at_least(required_index_buffer_size);
                self.index_buffer.buffer = create_index_buffer(device, self.index_buffer.capacity);
            }

            let index_buffer_staging = queue.write_buffer_with(
                &self.index_buffer.buffer,
                0,
                NonZeroU64::new(required_index_buffer_size).unwrap(),
            );

            let Some(mut index_buffer_staging) = index_buffer_staging else {
                panic!("Failed to create staging buffer for index data. Index count: {index_count}. Required index buffer size: {required_index_buffer_size}. Actual size {} and capacity: {} (bytes)", self.index_buffer.buffer.size(), self.index_buffer.capacity);
            };

            let mut index_offset = 0;
            for epaint::ClippedPrimitive { primitive, .. } in paint_jobs {
                match primitive {
                    Primitive::Mesh(mesh) => {
                        let size = mesh.indices.len() * std::mem::size_of::<u32>();
                        let slice = index_offset..(size + index_offset);
                        index_buffer_staging[slice.clone()]
                            .copy_from_slice(bytemuck::cast_slice(&mesh.indices));
                        self.index_buffer.slices.push(slice);
                        index_offset += size;
                    }
                    Primitive::Callback(_) => {}
                }
            }
        }
        if vertex_count > 0 {
            crate::profile_scope!("vertices", vertex_count.to_string());

            self.vertex_buffer.slices.clear();

            let required_vertex_buffer_size = (std::mem::size_of::<Vertex>() * vertex_count) as u64;
            if self.vertex_buffer.capacity < required_vertex_buffer_size {
                // Resize vertex buffer if needed.
                self.vertex_buffer.capacity =
                    (self.vertex_buffer.capacity * 2).at_least(required_vertex_buffer_size);
                self.vertex_buffer.buffer =
                    create_vertex_buffer(device, self.vertex_buffer.capacity);
            }

            let vertex_buffer_staging = queue.write_buffer_with(
                &self.vertex_buffer.buffer,
                0,
                NonZeroU64::new(required_vertex_buffer_size).unwrap(),
            );

            let Some(mut vertex_buffer_staging) = vertex_buffer_staging else {
                panic!("Failed to create staging buffer for vertex data. Vertex count: {vertex_count}. Required vertex buffer size: {required_vertex_buffer_size}. Actual size {} and capacity: {} (bytes)", self.vertex_buffer.buffer.size(), self.vertex_buffer.capacity);
            };

            let mut vertex_offset = 0;
            for epaint::ClippedPrimitive { primitive, .. } in paint_jobs {
                match primitive {
                    Primitive::Mesh(mesh) => {
                        let size = mesh.vertices.len() * std::mem::size_of::<Vertex>();
                        let slice = vertex_offset..(size + vertex_offset);
                        vertex_buffer_staging[slice.clone()]
                            .copy_from_slice(bytemuck::cast_slice(&mesh.vertices));
                        self.vertex_buffer.slices.push(slice);
                        vertex_offset += size;
                    }
                    Primitive::Callback(_) => {}
                }
            }
        }

        let mut user_cmd_bufs = Vec::new();
        {
            crate::profile_scope!("prepare callbacks");
            for callback in &callbacks {
                user_cmd_bufs.extend(callback.prepare(
                    device,
                    queue,
                    screen_descriptor,
                    encoder,
                    &mut self.callback_resources,
                ));
            }
        }
        {
            crate::profile_scope!("finish prepare callbacks");
            for callback in &callbacks {
                user_cmd_bufs.extend(callback.finish_prepare(
                    device,
                    queue,
                    encoder,
                    &mut self.callback_resources,
                ));
            }
        }

        user_cmd_bufs
    }
}

fn create_sampler(
    options: epaint::textures::TextureOptions,
    device: &wgpu::Device,
) -> wgpu::Sampler {
    let mag_filter = match options.magnification {
        epaint::textures::TextureFilter::Nearest => wgpu::FilterMode::Nearest,
        epaint::textures::TextureFilter::Linear => wgpu::FilterMode::Linear,
    };
    let min_filter = match options.minification {
        epaint::textures::TextureFilter::Nearest => wgpu::FilterMode::Nearest,
        epaint::textures::TextureFilter::Linear => wgpu::FilterMode::Linear,
    };
    let address_mode = match options.wrap_mode {
        epaint::textures::TextureWrapMode::ClampToEdge => wgpu::AddressMode::ClampToEdge,
        epaint::textures::TextureWrapMode::Repeat => wgpu::AddressMode::Repeat,
        epaint::textures::TextureWrapMode::MirroredRepeat => wgpu::AddressMode::MirrorRepeat,
    };
    device.create_sampler(&wgpu::SamplerDescriptor {
        label: Some(&format!(
            "egui sampler (mag: {mag_filter:?}, min {min_filter:?})"
        )),
        mag_filter,
        min_filter,
        address_mode_u: address_mode,
        address_mode_v: address_mode,
        ..Default::default()
    })
}

fn create_vertex_buffer(device: &wgpu::Device, size: u64) -> wgpu::Buffer {
    crate::profile_function!();
    device.create_buffer(&wgpu::BufferDescriptor {
        label: Some("egui_vertex_buffer"),
        usage: wgpu::BufferUsages::VERTEX | wgpu::BufferUsages::COPY_DST,
        size,
        mapped_at_creation: false,
    })
}

fn create_index_buffer(device: &wgpu::Device, size: u64) -> wgpu::Buffer {
    crate::profile_function!();
    device.create_buffer(&wgpu::BufferDescriptor {
        label: Some("egui_index_buffer"),
        usage: wgpu::BufferUsages::INDEX | wgpu::BufferUsages::COPY_DST,
        size,
        mapped_at_creation: false,
    })
}

/// A Rect in physical pixel space, used for setting clipping rectangles.
struct ScissorRect {
    x: u32,
    y: u32,
    width: u32,
    height: u32,
}

impl ScissorRect {
    fn new(clip_rect: &epaint::Rect, pixels_per_point: f32, target_size: [u32; 2]) -> Self {
        // Transform clip rect to physical pixels:
        let clip_min_x = pixels_per_point * clip_rect.min.x;
        let clip_min_y = pixels_per_point * clip_rect.min.y;
        let clip_max_x = pixels_per_point * clip_rect.max.x;
        let clip_max_y = pixels_per_point * clip_rect.max.y;

        // Round to integer:
        let clip_min_x = clip_min_x.round() as u32;
        let clip_min_y = clip_min_y.round() as u32;
        let clip_max_x = clip_max_x.round() as u32;
        let clip_max_y = clip_max_y.round() as u32;

        // Clamp:
        let clip_min_x = clip_min_x.clamp(0, target_size[0]);
        let clip_min_y = clip_min_y.clamp(0, target_size[1]);
        let clip_max_x = clip_max_x.clamp(clip_min_x, target_size[0]);
        let clip_max_y = clip_max_y.clamp(clip_min_y, target_size[1]);

        Self {
            x: clip_min_x,
            y: clip_min_y,
            width: clip_max_x - clip_min_x,
            height: clip_max_y - clip_min_y,
        }
    }
}

#[test]
fn renderer_impl_send_sync() {
    fn assert_send_sync<T: Send + Sync>() {}
    assert_send_sync::<Renderer>();
}