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
// SPDX-License-Identifier: AGPL-3.0-or-later
//! Laplacian Stencil (7-point 3D)
//!
//! **Physics**: Finite difference approximation of ∇²u
//! **Use Case**: Diffusion, electrostatics (PPPM), wave equations
//!
//! **Deep Debt Compliance**:
//! - ✅ Pure WGSL shader
//! - ✅ Periodic boundaries
//! - ✅ Zero unsafe code
use crate::error::{BarracudaError, Result};
use crate::tensor::Tensor;
/// Laplacian stencil operation (7-point 3D)
///
/// Computes ∇²u using central difference formula.
/// Includes periodic boundary conditions.
pub struct Laplacian {
field: Tensor, // [nx, ny, nz] input field
grid_spacing: f32, // h (mesh spacing)
}
impl Laplacian {
/// Creates a Laplacian stencil for a 3D field with given grid spacing.
///
/// # Errors
///
/// Returns [`Err`] if field is not 3D or `grid_spacing` is not positive.
pub fn new(field: Tensor, grid_spacing: f32) -> Result<Self> {
let shape = field.shape();
if shape.len() != 3 {
return Err(BarracudaError::InvalidShape {
expected: vec![0, 0, 0],
actual: shape.to_vec(),
});
}
if grid_spacing <= 0.0 {
return Err(BarracudaError::Device(
"Grid spacing h must be positive".to_string(),
));
}
Ok(Self {
field,
grid_spacing,
})
}
/// Execute Laplacian calculation
///
/// # Returns
/// Laplacian field [nx, ny, nz]
///
/// # Errors
///
/// Returns [`Err`] if buffer allocation, GPU dispatch, or buffer
/// readback fails (e.g. device lost or out of memory).
pub fn execute(self) -> Result<Tensor> {
let device = self.field.device();
let shape = self.field.shape();
let (nx, ny, nz) = (shape[0], shape[1], shape[2]);
// Create output buffer
let output_size = (nx * ny * nz * std::mem::size_of::<f32>()) as u64;
let output_buffer = device.device.create_buffer(&wgpu::BufferDescriptor {
label: Some("Laplacian Output"),
size: output_size,
usage: wgpu::BufferUsages::STORAGE | wgpu::BufferUsages::COPY_SRC,
mapped_at_creation: false,
});
#[repr(C)]
#[derive(Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
struct Params {
nx: u32,
ny: u32,
nz: u32,
h_squared: f32,
}
let params = Params {
nx: nx as u32,
ny: ny as u32,
nz: nz as u32,
h_squared: self.grid_spacing * self.grid_spacing,
};
let params_buffer = device
.device
.create_buffer_init(&wgpu::util::BufferInitDescriptor {
label: Some("Laplacian Params"),
contents: bytemuck::bytes_of(¶ms),
usage: wgpu::BufferUsages::UNIFORM,
});
let shader =
device.compile_shader(include_str!("laplacian.wgsl"), Some("Laplacian Shader"));
let bind_group_layout =
device
.device
.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
label: Some("Laplacian BGL"),
entries: &[
wgpu::BindGroupLayoutEntry {
binding: 0,
visibility: wgpu::ShaderStages::COMPUTE,
ty: wgpu::BindingType::Buffer {
ty: wgpu::BufferBindingType::Storage { read_only: true },
has_dynamic_offset: false,
min_binding_size: None,
},
count: None,
},
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,
},
wgpu::BindGroupLayoutEntry {
binding: 2,
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
.device
.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
label: Some("Laplacian PL"),
bind_group_layouts: &[&bind_group_layout],
immediate_size: 0,
});
let pipeline = device
.device
.create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
label: Some("Laplacian Pipeline"),
layout: Some(&pipeline_layout),
module: &shader,
entry_point: Some("main"),
cache: None,
compilation_options: Default::default(),
});
let bind_group = device.device.create_bind_group(&wgpu::BindGroupDescriptor {
label: Some("Laplacian BG"),
layout: &bind_group_layout,
entries: &[
wgpu::BindGroupEntry {
binding: 0,
resource: self.field.buffer().as_entire_binding(),
},
wgpu::BindGroupEntry {
binding: 1,
resource: output_buffer.as_entire_binding(),
},
wgpu::BindGroupEntry {
binding: 2,
resource: params_buffer.as_entire_binding(),
},
],
});
let mut encoder = device.create_encoder_guarded(&wgpu::CommandEncoderDescriptor {
label: Some("Laplacian Encoder"),
});
{
let mut pass = encoder.begin_compute_pass(&wgpu::ComputePassDescriptor {
label: Some("Laplacian Pass"),
timestamp_writes: None,
});
pass.set_pipeline(&pipeline);
pass.set_bind_group(0, Some(&bind_group), &[]);
// 3D workgroup dispatch (4x4x4 threads per workgroup, 64 total)
let workgroups_x = (nx as u32).div_ceil(4);
let workgroups_y = (ny as u32).div_ceil(4);
let workgroups_z = (nz as u32).div_ceil(4);
pass.dispatch_workgroups(workgroups_x, workgroups_y, workgroups_z);
}
device.submit_commands(Some(encoder.finish()));
Ok(Tensor::from_buffer(
output_buffer,
vec![nx, ny, nz],
device.clone(),
))
}
}
#[cfg(test)]
mod tests {
use super::*;
#[tokio::test]
// Test un-ignored - issue was test code structure, not tensor implementation
async fn test_laplacian_simple() {
let device = crate::device::test_pool::get_test_device().await;
// Simple 3x3x3 grid
let (nx, ny, nz) = (3, 3, 3);
let size = nx * ny * nz;
// Set all values to same number, Laplacian should be zero everywhere
let data = vec![1.0f32; size];
let field_tensor = Tensor::from_data(&data, vec![nx, ny, nz], device).unwrap();
// Verify input (explicit validation to prevent rustc optimization issues)
let field_check = field_tensor.to_vec().unwrap();
assert_eq!(field_check.len(), size, "Field size mismatch");
// All values should be 1.0
for (i, &val) in field_check.iter().enumerate() {
assert_eq!(
val, 1.0,
"Input corrupted at index {i}: expected 1.0, got {val}"
);
}
let laplacian = Laplacian::new(field_tensor, 1.0).unwrap();
let result = laplacian.execute().unwrap();
let lap_data = result.to_vec().unwrap();
// For constant field, Laplacian should be zero everywhere
// ∇²(constant) = 0
for (i, &val) in lap_data.iter().enumerate() {
assert!(
val.abs() < 1e-5,
"Index {i} Laplacian should be ~0, got {val}"
);
}
}
}