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
// SPDX-License-Identifier: AGPL-3.0-or-later
//! FHE Coefficient Extraction Operation
//!
//! **Purpose**: Extract specific polynomial coefficients from ciphertext
//!
//! **Algorithm**: Masked coefficient selection
//! - Apply zero-mask except at target index
//! - Extract single coefficient value
//! - Useful for slot-based decryption (CKKS)
//!
//! **Deep Debt Compliance**:
//! - ✅ Pure Rust + WGSL (no unsafe)
//! - ✅ GPU-accelerated (parallel masking)
//! - ✅ Numerically exact (no approximation)
//! - ✅ Production-ready (bounds checking)
//!
//! ## Use Cases
//!
//! 1. **Selective Decryption**: Decrypt single slot without full decryption
//! 2. **Debugging**: Extract specific coefficients for validation
//! 3. **CKKS Slots**: Access individual encrypted values in SIMD packing
//! 4. **Noise Analysis**: Examine specific coefficient noise levels
//!
//! ## Mathematical Background
//!
//! For a polynomial ciphertext:
//! ```text
//! ct(X) = c_0 + c_1*X + c_2*X² + ... + c_{n-1}*X^{n-1}
//! ```
//!
//! Extraction creates masked polynomial:
//! ```text
//! extract(ct, i) = c_i*X^i (all other coefficients = 0)
//! ```
//!
//! After decryption, only the i-th plaintext slot is visible.
//!
//! ## Usage Example
//!
//! ```rust,ignore
//! use barracuda::ops::fhe_extract::FheExtract;
//! use barracuda::prelude::Tensor;
//!
//! # async fn example() -> Result<(), Box<dyn std::error::Error>> {
//! // Ciphertext polynomial (degree 4096)
//! let ct = Tensor::from_u64_poly(&ciphertext, 4096).await?;
//!
//! // Extract coefficient at index 42
//! let extract_op = FheExtract::new(ct, 4096, 42)?;
//! let ct_masked = extract_op.execute()?;
//!
//! // ct_masked has only coefficient 42 non-zero
//! // Decrypt to get plaintext[42]
//! # Ok(())
//! # }
//! ```
use crate::device::{DeviceCapabilities, WorkloadType};
use crate::error::{BarracudaError, Result};
use crate::tensor::Tensor;
/// FHE Coefficient Extraction operation
///
/// Masks all coefficients except the target index.
pub struct FheExtract {
input: Tensor,
degree: u32,
index: u32,
pipeline: wgpu::ComputePipeline,
bind_group_layout: wgpu::BindGroupLayout,
}
impl FheExtract {
/// Create a new coefficient extraction operation
/// **Parameters**:
/// - `input`: Ciphertext polynomial (2*degree u32 values, u64 emulated)
/// - `degree`: Polynomial degree (power of 2)
/// - `index`: Coefficient index to extract (0 <= index < degree)
/// **Returns**: `FheExtract` operation ready to execute
/// **Errors**:
/// - Invalid degree (not power of 2)
/// - Index out of bounds (>= degree)
/// - Input tensor size mismatch
/// # Errors
/// Returns [`Err`] if degree is not a power of 2, index >= degree, or input
/// size does not match degree.
pub fn new(input: Tensor, degree: u32, index: u32) -> Result<Self> {
// ✅ VALIDATION: Degree must be power of 2
if !degree.is_power_of_two() || degree < 4 {
return Err(BarracudaError::InvalidInput {
message: format!("Degree must be power of 2 >= 4, got {degree}"),
});
}
// ✅ VALIDATION: Index must be in bounds
if index >= degree {
return Err(BarracudaError::InvalidInput {
message: format!("Index {index} out of bounds for degree {degree}"),
});
}
// ✅ VALIDATION: Input tensor must be 2*degree (u64 as 2xu32)
let expected_size = (degree * 2) as usize;
if input.shape()[0] != expected_size {
return Err(BarracudaError::InvalidInput {
message: format!(
"Input must have {} elements (degree={}, u64 emulated), got {}",
expected_size,
degree,
input.shape()[0]
),
});
}
let device = input.device();
// Load WGSL shader
let shader_source = include_str!("fhe_extract.wgsl");
let shader_module = device.compile_shader(shader_source, Some("FHE Extract Shader"));
// Create bind group layout
let bind_group_layout =
device
.device
.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
label: Some("FHE Extract Bind Group Layout"),
entries: &[
// Input buffer
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,
},
// Output buffer
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,
},
// Parameters buffer
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,
},
],
});
// Create pipeline
let pipeline_layout =
device
.device
.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
label: Some("FHE Extract Pipeline Layout"),
bind_group_layouts: &[&bind_group_layout],
immediate_size: 0,
});
let pipeline = device
.device
.create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
label: Some("FHE Extract Pipeline"),
layout: Some(&pipeline_layout),
module: &shader_module,
entry_point: Some("extract_coefficient"),
cache: None,
compilation_options: Default::default(),
});
Ok(Self {
input,
degree,
index,
pipeline,
bind_group_layout,
})
}
/// Execute coefficient extraction on GPU
/// **Returns**: Tensor with all coefficients zero except at target index
/// **Performance**: O(n) GPU parallel execution
/// # 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.input.device();
// Create output buffer
let output_buffer = device.device.create_buffer(&wgpu::BufferDescriptor {
label: Some("FHE Extract Output"),
size: self.input.buffer().size(),
usage: wgpu::BufferUsages::STORAGE | wgpu::BufferUsages::COPY_SRC,
mapped_at_creation: false,
});
// Create parameters buffer
#[repr(C)]
#[derive(Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
struct ExtractParams {
degree: u32,
target_index: u32,
_padding: [u32; 2], // Align to 16 bytes
}
let params = ExtractParams {
degree: self.degree,
target_index: self.index,
_padding: [0; 2],
};
let params_buffer = device
.device
.create_buffer_init(&wgpu::util::BufferInitDescriptor {
label: Some("FHE Extract Params"),
contents: bytemuck::bytes_of(¶ms),
usage: wgpu::BufferUsages::UNIFORM,
});
// Create bind group
let bind_group = device.device.create_bind_group(&wgpu::BindGroupDescriptor {
label: Some("FHE Extract Bind Group"),
layout: &self.bind_group_layout,
entries: &[
wgpu::BindGroupEntry {
binding: 0,
resource: self.input.buffer().as_entire_binding(),
},
wgpu::BindGroupEntry {
binding: 1,
resource: output_buffer.as_entire_binding(),
},
wgpu::BindGroupEntry {
binding: 2,
resource: params_buffer.as_entire_binding(),
},
],
});
// ✅ GPU EXECUTION: Parallel coefficient masking
let mut encoder = device.create_encoder_guarded(&wgpu::CommandEncoderDescriptor {
label: Some("FHE Extract Encoder"),
});
{
let mut compute_pass = encoder.begin_compute_pass(&wgpu::ComputePassDescriptor {
label: Some("FHE Extract Pass"),
timestamp_writes: None,
});
compute_pass.set_pipeline(&self.pipeline);
compute_pass.set_bind_group(0, Some(&bind_group), &[]);
// Deep Debt Evolution: Capability-based dispatch
let caps = DeviceCapabilities::from_device(device);
let optimal_wg_size = caps.optimal_workgroup_size(WorkloadType::FHE);
let workgroups = self.degree.div_ceil(optimal_wg_size);
compute_pass.dispatch_workgroups(workgroups, 1, 1);
}
device.submit_commands(std::iter::once(encoder.finish()));
// Return result tensor
Ok(Tensor::from_buffer(
output_buffer,
self.input.shape().to_vec(),
device.clone(),
))
}
}
#[cfg(test)]
mod tests {
use super::*;
#[tokio::test]
async fn test_extract_validation() {
let device = crate::device::test_pool::get_test_device().await;
let t = || Tensor::zeros_on(vec![8], device.clone());
// Test invalid degree
let result = FheExtract::new(t().await.unwrap(), 3, 0);
assert!(result.is_err());
// Test index out of bounds
let result = FheExtract::new(t().await.unwrap(), 4, 4);
assert!(result.is_err());
// Test index >= degree
let result = FheExtract::new(t().await.unwrap(), 4, 5);
assert!(result.is_err());
}
// Full integration tests in tests/fhe_shader_unit_tests.rs
}