ringkernel-wgpu 0.4.2

WebGPU backend for RingKernel - cross-platform GPU support
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

//! Shader compilation and management.

use ringkernel_core::error::Result;

use crate::adapter::WgpuAdapter;

/// Compiled compute pipeline.
pub struct ComputePipeline {
    /// The wgpu pipeline.
    pipeline: wgpu::ComputePipeline,
    /// Bind group layout.
    bind_group_layout: wgpu::BindGroupLayout,
    /// Pipeline layout.
    #[allow(dead_code)]
    pipeline_layout: wgpu::PipelineLayout,
    /// Workgroup size.
    #[allow(dead_code)]
    workgroup_size: (u32, u32, u32),
}

impl ComputePipeline {
    /// Create a new compute pipeline from WGSL source.
    pub fn new(
        adapter: &WgpuAdapter,
        wgsl_source: &str,
        entry_point: &str,
        workgroup_size: (u32, u32, u32),
    ) -> Result<Self> {
        let device = adapter.device();

        // Create shader module
        let shader_module = device.create_shader_module(wgpu::ShaderModuleDescriptor {
            label: Some("RingKernel Shader"),
            source: wgpu::ShaderSource::Wgsl(wgsl_source.into()),
        });

        // Create bind group layout for control block and queues
        let bind_group_layout = device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
            label: Some("RingKernel Bind Group Layout"),
            entries: &[
                // Control block
                wgpu::BindGroupLayoutEntry {
                    binding: 0,
                    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,
                },
                // Input queue
                wgpu::BindGroupLayoutEntry {
                    binding: 1,
                    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,
                },
                // Output queue
                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,
                },
            ],
        });

        let pipeline_layout = device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
            label: Some("RingKernel Pipeline Layout"),
            bind_group_layouts: &[&bind_group_layout],
            push_constant_ranges: &[],
        });

        let pipeline = device.create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
            label: Some("RingKernel Compute Pipeline"),
            layout: Some(&pipeline_layout),
            module: &shader_module,
            entry_point,
        });

        Ok(Self {
            pipeline,
            bind_group_layout,
            pipeline_layout,
            workgroup_size,
        })
    }

    /// Get the pipeline.
    pub fn pipeline(&self) -> &wgpu::ComputePipeline {
        &self.pipeline
    }

    /// Get bind group layout.
    pub fn bind_group_layout(&self) -> &wgpu::BindGroupLayout {
        &self.bind_group_layout
    }

    /// Get workgroup size.
    #[allow(dead_code)]
    pub fn workgroup_size(&self) -> (u32, u32, u32) {
        self.workgroup_size
    }
}

/// Create a bind group for the kernel.
pub fn create_bind_group(
    device: &wgpu::Device,
    layout: &wgpu::BindGroupLayout,
    control_block: wgpu::BindingResource,
    input_queue: wgpu::BindingResource,
    output_queue: wgpu::BindingResource,
) -> wgpu::BindGroup {
    device.create_bind_group(&wgpu::BindGroupDescriptor {
        label: Some("RingKernel Bind Group"),
        layout,
        entries: &[
            wgpu::BindGroupEntry {
                binding: 0,
                resource: control_block,
            },
            wgpu::BindGroupEntry {
                binding: 1,
                resource: input_queue,
            },
            wgpu::BindGroupEntry {
                binding: 2,
                resource: output_queue,
            },
        ],
    })
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::RING_KERNEL_WGSL_TEMPLATE;

    #[tokio::test]
    #[ignore] // May not have GPU in CI
    async fn test_pipeline_creation() {
        let adapter = WgpuAdapter::new().await.unwrap();
        let pipeline =
            ComputePipeline::new(&adapter, RING_KERNEL_WGSL_TEMPLATE, "main", (256, 1, 1)).unwrap();

        assert_eq!(pipeline.workgroup_size(), (256, 1, 1));
    }
}