image-convolution 0.1.0

Parallel image convolution on GPU.
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

use std::borrow::Cow;
use wgpu::util::DeviceExt;

pub struct GpuDevice {
    pub(crate) device: wgpu::Device,
    pub(crate) queue: wgpu::Queue,
}

pub fn create_gpu_device() -> GpuDevice {
    let (device, queue) = futures::executor::block_on(create_device_queue());
    GpuDevice { device, queue }
}

pub async fn create_device_queue() -> (wgpu::Device, wgpu::Queue) {
    let instance = wgpu::Instance::new(wgpu::BackendBit::PRIMARY);
    let adapter = instance
        .request_adapter(&wgpu::RequestAdapterOptions {
            power_preference: wgpu::PowerPreference::HighPerformance,
            compatible_surface: None,
        })
        .await
        .expect("Failed to find an appropriate adapter");

    adapter
        .request_device(
            &wgpu::DeviceDescriptor {
                label: None,
                features: wgpu::Features::empty(),
                limits: wgpu::Limits::default(),
            },
            None,
        )
        .await
        .expect("Failed to create device")
}

impl GpuDevice {
    pub fn create_buffer(&self, label: &str, size: u64) -> wgpu::Buffer {
        self.device.create_buffer(&wgpu::BufferDescriptor {
            label: Some(label),
            size,
            usage: wgpu::BufferUsage::STORAGE | wgpu::BufferUsage::COPY_SRC,
            mapped_at_creation: false,
        })
    }

    pub fn create_data_buffer(&self, label: &str, contents: &[u8]) -> wgpu::Buffer {
        self.device
            .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                label: Some(label),
                contents,
                usage: wgpu::BufferUsage::STORAGE | wgpu::BufferUsage::COPY_SRC,
            })
    }

    pub fn create_uniform_buffer(&self, label: &str, contents: &[u8]) -> wgpu::Buffer {
        self.device
            .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                label: Some(label),
                contents,
                usage: wgpu::BufferUsage::UNIFORM,
            })
    }

    pub fn create_output_buffer(&self, label: &str, size: u64) -> wgpu::Buffer {
        self.device.create_buffer(&wgpu::BufferDescriptor {
            label: Some(label),
            size,
            usage: wgpu::BufferUsage::MAP_READ | wgpu::BufferUsage::COPY_DST,
            mapped_at_creation: false,
        })
    }

    pub fn create_bind_group(
        &self,
        buffers: &[(&wgpu::Buffer, u64, wgpu::BufferBindingType)],
    ) -> (wgpu::BindGroupLayout, wgpu::BindGroup) {
        let layout_entries = buffers
            .iter()
            .enumerate()
            .map(|(index, (_, size, ty))| wgpu::BindGroupLayoutEntry {
                binding: index as u32,
                visibility: wgpu::ShaderStage::COMPUTE,
                ty: wgpu::BindingType::Buffer {
                    ty: *ty,
                    has_dynamic_offset: false,
                    min_binding_size: wgpu::BufferSize::new(*size),
                },
                count: None,
            })
            .collect::<Vec<_>>();
        let bind_group_layout =
            self.device
                .create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
                    label: None,
                    entries: &layout_entries,
                });
        let group_entries = buffers
            .iter()
            .enumerate()
            .map(|(index, (buffer, _, _))| wgpu::BindGroupEntry {
                binding: index as u32,
                resource: buffer.as_entire_binding(),
            })
            .collect::<Vec<_>>();
        let bind_group = self.device.create_bind_group(&wgpu::BindGroupDescriptor {
            label: None,
            layout: &bind_group_layout,
            entries: &group_entries,
        });
        (bind_group_layout, bind_group)
    }

    pub fn create_compute_pipeline(
        &self,
        buffers: &[(&wgpu::Buffer, u64, wgpu::BufferBindingType)],
        shader: &str,
    ) -> (wgpu::BindGroup, wgpu::ComputePipeline) {
        let (bind_group_layout, bind_group) = self.create_bind_group(buffers);

        // create shader module
        let cs_module = self
            .device
            .create_shader_module(&wgpu::ShaderModuleDescriptor {
                label: None,
                source: wgpu::ShaderSource::Wgsl(Cow::Borrowed(shader)),
                flags: wgpu::ShaderFlags::VALIDATION,
            });

        // create pipeline for shader
        let pipeline_layout = self
            .device
            .create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
                label: None,
                bind_group_layouts: &[&bind_group_layout],
                push_constant_ranges: &[],
            });
        let compute_pipeline =
            self.device
                .create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
                    label: None,
                    layout: Some(&pipeline_layout),
                    module: &cs_module,
                    entry_point: "main",
                });

        (bind_group, compute_pipeline)
    }
}