primal-barracuda 0.4.0

barraCuda — BARrier-free Rust Abstracted Cross-platform Unified Dimensional Algebra. Vendor-agnostic GPU compute via WGSL.
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
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
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
// SPDX-License-Identifier: AGPL-3.0-or-later
//! Batched Bisection Root-Finding (GPU) — Parallel Root-Finding
//!
//! Solves many independent root-finding problems in parallel on GPU.
//! Each problem runs bisection concurrently with full f64 precision.
//!
//! **Use cases**:
//! - BCS pairing: Find μ where Σ `v²_k(μ)` = N for each nucleus
//! - Multi-system parameter fitting
//! - Batch chemical equilibrium calculations
//!
//! **Deep Debt Principles**:
//! - Pure WGSL implementation (hardware-agnostic)
//! - Full f64 precision
//! - Safe Rust wrapper (no unsafe code)
//!
//! # Example
//!
//! ```rust,ignore
//! let device = WgpuDevice::new().await?;
//! let bisect = BatchedBisectionGpu::new(device.clone(), 64, 1e-12)?;
//!
//! // Find √2, √3, √5 in parallel
//! let lower = vec![0.0, 0.0, 0.0];
//! let upper = vec![2.0, 2.0, 3.0];
//! let targets = vec![2.0, 3.0, 5.0]; // x² = target
//!
//! let roots = bisect.solve_polynomial(&lower, &upper, &targets)?;
//! // roots ≈ [1.414, 1.732, 2.236]
//! ```

use crate::device::WgpuDevice;
use crate::device::capabilities::WORKGROUP_SIZE_COMPACT;
use crate::device::compute_pipeline::ComputeDispatch;
use crate::error::{BarracudaError, Result};
use bytemuck::{Pod, Zeroable};
use std::sync::Arc;

/// Parameters for batched bisection shader
#[repr(C)]
#[derive(Copy, Clone, Pod, Zeroable)]
struct BisectionParams {
    batch_size: u32,
    max_iterations: u32,
    n_levels: u32,
    use_degeneracy: u32, // 1 = use deg_k weights, 0 = assume deg_k=1
    tolerance_lo: u32,   // f64 as two u32s for alignment
    tolerance_hi: u32,
}

impl BisectionParams {
    fn new(
        batch_size: u32,
        max_iterations: u32,
        n_levels: u32,
        use_degeneracy: bool,
        tolerance: f64,
    ) -> Self {
        let tol_bits = tolerance.to_bits();
        Self {
            batch_size,
            max_iterations,
            n_levels,
            use_degeneracy: u32::from(use_degeneracy),
            tolerance_lo: tol_bits as u32,
            tolerance_hi: (tol_bits >> 32) as u32,
        }
    }
}

/// GPU-accelerated batched bisection root-finding
///
/// Solves N independent bisection problems in parallel.
pub struct BatchedBisectionGpu {
    device: Arc<WgpuDevice>,
    max_iterations: u32,
    tolerance: f64,
}

/// Result of batched bisection
pub struct BisectionResult {
    /// Found roots [`batch_size`]
    pub roots: Vec<f64>,
    /// Number of iterations used per problem [`batch_size`]
    pub iterations: Vec<u32>,
}

impl BatchedBisectionGpu {
    /// Create a new batched bisection solver
    /// # Arguments
    /// * `device` - `WgpuDevice`
    /// * `max_iterations` - Maximum bisection iterations per problem (typically 50-100)
    /// * `tolerance` - Convergence tolerance (typically 1e-10 to 1e-14)
    /// # Errors
    /// Returns [`Err`] if `tolerance` ≤ 0.
    pub fn new(device: Arc<WgpuDevice>, max_iterations: u32, tolerance: f64) -> Result<Self> {
        if tolerance <= 0.0 {
            return Err(BarracudaError::InvalidInput {
                message: "Tolerance must be positive".to_string(),
            });
        }
        Ok(Self {
            device,
            max_iterations,
            tolerance,
        })
    }

    fn wgsl_shader() -> &'static str {
        include_str!("../shaders/optimizer/batched_bisection_f64.wgsl")
    }

    /// Solve batched polynomial root-finding: x² = target
    /// Finds √target for each problem in parallel.
    /// This is a validation/test function.
    /// # Arguments
    /// * `lower` - Lower bounds [`batch_size`]
    /// * `upper` - Upper bounds [`batch_size`]
    /// * `targets` - Target values (find x where x² = target) [`batch_size`]
    /// # Errors
    /// Returns [`Err`] if `lower`, `upper`, and `targets` lengths do not match, or if GPU
    /// execution or buffer readback fails.
    pub fn solve_polynomial(
        &self,
        lower: &[f64],
        upper: &[f64],
        targets: &[f64],
    ) -> Result<BisectionResult> {
        if lower.len() != upper.len() || lower.len() != targets.len() {
            return Err(BarracudaError::InvalidInput {
                message: format!(
                    "Array lengths must match: lower={}, upper={}, targets={}",
                    lower.len(),
                    upper.len(),
                    targets.len()
                ),
            });
        }

        self.solve_internal(lower, upper, targets, 1, false, "batched_bisection_poly")
    }

    /// Solve BCS particle number equation: Σ `v²_k(μ)` = N
    /// Finds chemical potential μ for each nucleus such that the
    /// BCS occupation numbers sum to the target particle number.
    /// # Arguments
    /// * `lower` - Lower bounds for μ [`batch_size`]
    /// * `upper` - Upper bounds for μ [`batch_size`]
    /// * `eigenvalues` - Single-particle energies [`batch_size`, `n_levels`]
    /// * `delta` - Pairing gap for each problem [`batch_size`]
    /// * `target_n` - Target particle number for each problem [`batch_size`]
    /// # Errors
    /// Returns [`Err`] if array lengths do not match or `eigenvalues.len()` is not divisible by
    /// `batch_size`, or if GPU execution or buffer readback fails.
    pub fn solve_bcs(
        &self,
        lower: &[f64],
        upper: &[f64],
        eigenvalues: &[f64],
        delta: &[f64],
        target_n: &[f64],
    ) -> Result<BisectionResult> {
        let batch_size = lower.len();
        if upper.len() != batch_size || delta.len() != batch_size || target_n.len() != batch_size {
            return Err(BarracudaError::InvalidInput {
                message: "Array lengths must match batch_size".to_string(),
            });
        }

        // Calculate n_levels
        if !eigenvalues.len().is_multiple_of(batch_size) {
            return Err(BarracudaError::InvalidInput {
                message: format!(
                    "eigenvalues length {} must be divisible by batch_size {}",
                    eigenvalues.len(),
                    batch_size
                ),
            });
        }
        let n_levels = eigenvalues.len() / batch_size;

        // Pack params: [ε_0, ..., ε_{n-1}, Δ, N] per problem
        let params_per_problem = n_levels + 2;
        let mut params = Vec::with_capacity(batch_size * params_per_problem);
        for i in 0..batch_size {
            // Eigenvalues for this problem
            for j in 0..n_levels {
                params.push(eigenvalues[i * n_levels + j]);
            }
            // Delta and target N
            params.push(delta[i]);
            params.push(target_n[i]);
        }

        self.solve_internal(
            lower,
            upper,
            &params,
            n_levels as u32,
            false,
            "batched_bisection",
        )
    }

    /// Solve BCS pairing equations with level degeneracy (`deg_k`)
    /// For nuclear HFB: `deg_k` = 2j+1 (spin degeneracy of each level)
    /// **Formula**: Find μ such that `Σ_k` `deg_k` · `v²_k(μ)` = N
    /// # Arguments
    /// * `lower` - Lower bounds for μ [`batch_size`]
    /// * `upper` - Upper bounds for μ [`batch_size`]
    /// * `eigenvalues` - Packed energy levels [`batch_size` × `n_levels`]
    /// * `degeneracies` - Degeneracy of each level [`batch_size` × `n_levels`]
    /// * `delta` - BCS pairing gap per problem [`batch_size`]
    /// * `target_n` - Target particle number per problem [`batch_size`]
    /// # Evolution
    /// Added Feb 16, 2026 per hotSpring handoff TIER 3.1
    /// # Errors
    /// Returns [`Err`] if array lengths do not match, `eigenvalues.len()` is not divisible by
    /// `batch_size`, or `degeneracies.len() != eigenvalues.len()`, or if GPU execution or
    /// buffer readback fails.
    pub fn solve_bcs_with_degeneracy(
        &self,
        lower: &[f64],
        upper: &[f64],
        eigenvalues: &[f64],
        degeneracies: &[f64],
        delta: &[f64],
        target_n: &[f64],
    ) -> Result<BisectionResult> {
        let batch_size = lower.len();
        if upper.len() != batch_size || delta.len() != batch_size || target_n.len() != batch_size {
            return Err(BarracudaError::InvalidInput {
                message: "Array lengths must match batch_size".to_string(),
            });
        }

        // Calculate n_levels
        if !eigenvalues.len().is_multiple_of(batch_size) {
            return Err(BarracudaError::InvalidInput {
                message: format!(
                    "eigenvalues length {} must be divisible by batch_size {}",
                    eigenvalues.len(),
                    batch_size
                ),
            });
        }
        let n_levels = eigenvalues.len() / batch_size;

        if degeneracies.len() != eigenvalues.len() {
            return Err(BarracudaError::InvalidInput {
                message: format!(
                    "degeneracies length {} must match eigenvalues length {}",
                    degeneracies.len(),
                    eigenvalues.len()
                ),
            });
        }

        // Pack params: [ε_0, ..., ε_{n-1}, deg_0, ..., deg_{n-1}, Δ, N] per problem
        let params_per_problem = n_levels * 2 + 2;
        let mut params = Vec::with_capacity(batch_size * params_per_problem);
        for i in 0..batch_size {
            // Eigenvalues for this problem
            for j in 0..n_levels {
                params.push(eigenvalues[i * n_levels + j]);
            }
            // Degeneracies for this problem
            for j in 0..n_levels {
                params.push(degeneracies[i * n_levels + j]);
            }
            // Delta and target N
            params.push(delta[i]);
            params.push(target_n[i]);
        }

        self.solve_internal(
            lower,
            upper,
            &params,
            n_levels as u32,
            true,
            "batched_bisection",
        )
    }

    fn solve_internal(
        &self,
        lower: &[f64],
        upper: &[f64],
        params: &[f64],
        n_levels: u32,
        use_degeneracy: bool,
        entry_point: &str,
    ) -> Result<BisectionResult> {
        let batch_size = lower.len();
        if batch_size == 0 {
            return Ok(BisectionResult {
                roots: vec![],
                iterations: vec![],
            });
        }

        // Create buffers
        let lower_bytes: &[u8] = bytemuck::cast_slice(lower);
        let lower_buffer =
            self.device
                .device
                .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                    label: Some("BatchedBisection lower"),
                    contents: lower_bytes,
                    usage: wgpu::BufferUsages::STORAGE,
                });

        let upper_bytes: &[u8] = bytemuck::cast_slice(upper);
        let upper_buffer =
            self.device
                .device
                .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                    label: Some("BatchedBisection upper"),
                    contents: upper_bytes,
                    usage: wgpu::BufferUsages::STORAGE,
                });

        let params_bytes: &[u8] = bytemuck::cast_slice(params);
        let params_buffer =
            self.device
                .device
                .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                    label: Some("BatchedBisection params"),
                    contents: params_bytes,
                    usage: wgpu::BufferUsages::STORAGE,
                });

        let roots_buffer = self.device.device.create_buffer(&wgpu::BufferDescriptor {
            label: Some("BatchedBisection roots"),
            size: (batch_size * 8) as u64,
            usage: wgpu::BufferUsages::STORAGE | wgpu::BufferUsages::COPY_SRC,
            mapped_at_creation: false,
        });

        let iterations_buffer = self.device.device.create_buffer(&wgpu::BufferDescriptor {
            label: Some("BatchedBisection iterations"),
            size: (batch_size * 4) as u64,
            usage: wgpu::BufferUsages::STORAGE | wgpu::BufferUsages::COPY_SRC,
            mapped_at_creation: false,
        });

        let config = BisectionParams::new(
            batch_size as u32,
            self.max_iterations,
            n_levels,
            use_degeneracy,
            self.tolerance,
        );
        let config_buffer = self
            .device
            .create_uniform_buffer("BatchedBisection config", &config);

        let n_workgroups = batch_size.div_ceil(WORKGROUP_SIZE_COMPACT as usize);
        ComputeDispatch::new(self.device.as_ref(), entry_point)
            .shader(Self::wgsl_shader(), entry_point)
            .f64()
            .storage_read(0, &lower_buffer)
            .storage_read(1, &upper_buffer)
            .storage_read(2, &params_buffer)
            .storage_rw(3, &roots_buffer)
            .storage_rw(4, &iterations_buffer)
            .uniform(5, &config_buffer)
            .dispatch(n_workgroups as u32, 1, 1)
            .submit()?;

        // Read back results
        let roots = self.device.read_f64_buffer(&roots_buffer, batch_size)?;
        let iterations = self.read_u32_buffer(&iterations_buffer, batch_size)?;

        Ok(BisectionResult { roots, iterations })
    }

    fn read_u32_buffer(&self, buffer: &wgpu::Buffer, count: usize) -> Result<Vec<u32>> {
        let staging = self.device.device.create_buffer(&wgpu::BufferDescriptor {
            label: Some("BatchedBisection u32 staging"),
            size: (count * 4) as u64,
            usage: wgpu::BufferUsages::MAP_READ | wgpu::BufferUsages::COPY_DST,
            mapped_at_creation: false,
        });

        let mut encoder = self
            .device
            .create_encoder_guarded(&wgpu::CommandEncoderDescriptor {
                label: Some("BatchedBisection u32 readback"),
            });
        encoder.copy_buffer_to_buffer(buffer, 0, &staging, 0, (count * 4) as u64);
        self.device.submit_commands(Some(encoder.finish()));

        let slice = staging.slice(..);
        let (sender, receiver) = std::sync::mpsc::channel();
        slice.map_async(wgpu::MapMode::Read, move |result| {
            // Channel send should not fail since receiver is alive during poll
            let _ = sender.send(result);
        });
        self.device.poll_safe()?;
        receiver
            .recv()
            .map_err(|_| BarracudaError::execution_failed("GPU buffer mapping channel closed"))?
            .map_err(|e| BarracudaError::execution_failed(e.to_string()))?;

        let data = slice.get_mapped_range();
        // chunks_exact(4) guarantees exactly 4-byte chunks
        let result: Vec<u32> = data
            .chunks_exact(4)
            .map(|chunk| {
                let arr: [u8; 4] = chunk.try_into().map_err(|_| {
                    BarracudaError::Internal("chunks_exact(4) invariant violated".to_string())
                })?;
                Ok(u32::from_le_bytes(arr))
            })
            .collect::<Result<Vec<_>>>()?;
        drop(data);
        staging.unmap();

        Ok(result)
    }
}

#[cfg(test)]
#[path = "batched_bisection_gpu_tests.rs"]
mod tests;