scena 1.0.2

A Rust-native scene-graph renderer with typed scene state, glTF assets, and explicit prepare/render lifecycles.
Documentation 
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

use super::output;
#[cfg(not(target_arch = "wasm32"))]
use super::pipeline::BYTES_PER_PIXEL;
use super::vertices::VERTEX_BYTE_LEN;

use super::super::RasterTarget;

#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
pub(in crate::render) struct GpuResourceStats {
    pub(in crate::render) buffers: u64,
    pub(in crate::render) textures: u64,
    pub(in crate::render) render_targets: u64,
    pub(in crate::render) pipelines: u64,
    pub(in crate::render) bind_groups: u64,
    pub(in crate::render) shader_modules: u64,
    pub(in crate::render) approximate_gpu_memory_bytes: u64,
    /// Plan line 778 commit 2: distinct material bind groups consumed by
    /// the unlit pass. Equals 1 when the renderer chose the batched
    /// `texture_2d_array<f32>` path (single shared bind group serviced via
    /// dynamic-offset uniforms) and equals the per-material slot count
    /// otherwise (one bind group per slot, including the synthetic
    /// fallback at index 0).
    pub(in crate::render) material_bind_groups: u32,
}

#[derive(Debug, Clone, Copy)]
pub(super) struct PreparedResourceEstimateInput {
    pub(super) target: RasterTarget,
    pub(super) vertex_count: usize,
    pub(super) has_surface_pipeline: bool,
    pub(super) shadow_maps: u64,
    pub(super) shadow_map_resolution: Option<u32>,
    pub(super) depth_prepass_passes: u64,
    pub(super) material_texture_count: u64,
    pub(super) material_texture_bytes: u64,
    /// Plan line 778 commit 2: distinct material bind groups in the
    /// prepared resource set. The estimator records this so the
    /// observable `RendererStats::material_bind_groups` reflects the
    /// actual GPU shape rather than the per-material count.
    pub(super) material_bind_groups: u32,
}

impl GpuResourceStats {
    pub(in crate::render) fn destruction_records(self) -> u64 {
        self.buffers
            + self.textures
            + self.render_targets
            + self.pipelines
            + self.bind_groups
            + self.shader_modules
    }
}

pub(super) fn estimate_prepared_resource_stats(
    input: PreparedResourceEstimateInput,
) -> GpuResourceStats {
    let PreparedResourceEstimateInput {
        target,
        vertex_count,
        has_surface_pipeline,
        shadow_maps,
        shadow_map_resolution,
        depth_prepass_passes,
        material_texture_count,
        material_texture_bytes,
        material_bind_groups,
    } = input;

    if vertex_count == 0 {
        return GpuResourceStats::default();
    }

    #[cfg(not(target_arch = "wasm32"))]
    let unpadded_bytes_per_row = target.width.saturating_mul(BYTES_PER_PIXEL);
    #[cfg(not(target_arch = "wasm32"))]
    let padded_bytes_per_row = align_to(unpadded_bytes_per_row, wgpu::COPY_BYTES_PER_ROW_ALIGNMENT);
    #[cfg(not(target_arch = "wasm32"))]
    let texture_bytes = u64::from(unpadded_bytes_per_row) * u64::from(target.height);
    #[cfg(not(target_arch = "wasm32"))]
    let readback_bytes = u64::from(padded_bytes_per_row) * u64::from(target.height);
    let vertex_bytes = (vertex_count * VERTEX_BYTE_LEN).max(4) as u64;
    let uniform_bytes = output::OUTPUT_UNIFORM_BYTE_LEN;
    let pipelines = 1 + u64::from(has_surface_pipeline) + depth_prepass_passes;
    #[cfg(not(target_arch = "wasm32"))]
    let shadow_map_bytes = shadow_map_resolution
        .map(|resolution| {
            let edge = u64::from(resolution);
            shadow_maps.saturating_mul(edge.saturating_mul(edge).saturating_mul(4))
        })
        .unwrap_or(0);
    #[cfg(target_arch = "wasm32")]
    let shadow_map_bytes = {
        let _ = shadow_map_resolution;
        let _ = shadow_maps;
        0
    };
    #[cfg(not(target_arch = "wasm32"))]
    let depth_prepass_bytes = u64::from(target.width)
        .saturating_mul(u64::from(target.height))
        .saturating_mul(4)
        .saturating_mul(depth_prepass_passes);
    #[cfg(target_arch = "wasm32")]
    let depth_prepass_bytes = {
        let _ = target;
        0
    };

    GpuResourceStats {
        #[cfg(not(target_arch = "wasm32"))]
        buffers: 3,
        #[cfg(target_arch = "wasm32")]
        buffers: 2,
        #[cfg(not(target_arch = "wasm32"))]
        textures: 1 + material_texture_count + shadow_maps + depth_prepass_passes,
        #[cfg(target_arch = "wasm32")]
        textures: material_texture_count,
        #[cfg(not(target_arch = "wasm32"))]
        render_targets: 1 + shadow_maps + depth_prepass_passes,
        #[cfg(target_arch = "wasm32")]
        render_targets: 1,
        pipelines,
        // Plan line 778 commit 2: the unlit pass binds 1 output bind group
        // + N material bind groups (1 when batched, slot count otherwise).
        // Adding `material_bind_groups` keeps the resource estimate
        // consistent with the actual GPU shape.
        bind_groups: 1 + u64::from(material_bind_groups),
        shader_modules: pipelines,
        material_bind_groups,
        #[cfg(not(target_arch = "wasm32"))]
        approximate_gpu_memory_bytes: texture_bytes
            + readback_bytes
            + vertex_bytes
            + uniform_bytes
            + material_texture_bytes
            + shadow_map_bytes
            + depth_prepass_bytes,
        #[cfg(target_arch = "wasm32")]
        approximate_gpu_memory_bytes: vertex_bytes
            + uniform_bytes
            + material_texture_bytes
            + shadow_map_bytes
            + depth_prepass_bytes,
    }
}

#[cfg(not(target_arch = "wasm32"))]
pub(super) fn align_to(value: u32, alignment: u32) -> u32 {
    value.div_ceil(alignment) * alignment
}

#[cfg(all(test, not(target_arch = "wasm32")))]
mod tests {
    use super::*;
    use crate::diagnostics::Backend;

    #[test]
    fn estimates_prepared_headless_gpu_resource_counters() {
        let target = RasterTarget {
            width: 4,
            height: 4,
            backend: Backend::HeadlessGpu,
        };

        let stats = estimate_prepared_resource_stats(estimate_input(target, 3));

        assert_eq!(stats.buffers, 3);
        assert_eq!(stats.textures, 2);
        assert_eq!(stats.render_targets, 1);
        assert_eq!(stats.pipelines, 1);
        assert_eq!(stats.bind_groups, 2);
        assert_eq!(stats.shader_modules, 1);
        assert_eq!(stats.destruction_records(), 10);
        assert!(stats.approximate_gpu_memory_bytes > 0);
    }

    #[test]
    fn estimates_empty_headless_gpu_resource_counters_at_baseline() {
        let target = RasterTarget {
            width: 4,
            height: 4,
            backend: Backend::HeadlessGpu,
        };

        let stats = estimate_prepared_resource_stats(estimate_input(target, 0));

        assert_eq!(stats, GpuResourceStats::default());
        assert_eq!(stats.destruction_records(), 0);
    }

    #[test]
    fn estimates_single_shadow_map_resource_counters() {
        let target = RasterTarget {
            width: 4,
            height: 4,
            backend: Backend::HeadlessGpu,
        };

        let stats = estimate_prepared_resource_stats(PreparedResourceEstimateInput {
            shadow_maps: 1,
            shadow_map_resolution: Some(2048),
            ..estimate_input(target, 3)
        });

        assert_eq!(stats.textures, 3);
        assert_eq!(stats.render_targets, 2);
        assert_eq!(stats.destruction_records(), 12);
        assert!(stats.approximate_gpu_memory_bytes >= 2048 * 2048 * 4);
    }

    #[test]
    fn estimates_depth_prepass_resource_counters() {
        let target = RasterTarget {
            width: 4,
            height: 4,
            backend: Backend::HeadlessGpu,
        };

        let stats = estimate_prepared_resource_stats(PreparedResourceEstimateInput {
            depth_prepass_passes: 1,
            ..estimate_input(target, 3)
        });

        assert_eq!(stats.textures, 3);
        assert_eq!(stats.render_targets, 2);
        assert_eq!(stats.pipelines, 2);
        assert_eq!(stats.shader_modules, 2);
        assert_eq!(stats.destruction_records(), 14);
        assert!(stats.approximate_gpu_memory_bytes >= 4 * 4 * 4);
    }

    fn estimate_input(target: RasterTarget, vertex_count: usize) -> PreparedResourceEstimateInput {
        PreparedResourceEstimateInput {
            target,
            vertex_count,
            has_surface_pipeline: false,
            shadow_maps: 0,
            shadow_map_resolution: None,
            depth_prepass_passes: 0,
            material_texture_count: 1,
            material_texture_bytes: 4,
            material_bind_groups: 1,
        }
    }
}