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
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
// SPDX-License-Identifier: AGPL-3.0-or-later
//! Generalized Symmetric Eigenvalue Problem (f64 CPU)
//!
//! Solves the generalized eigenvalue problem: Ax = λBx
//! where A is symmetric and B is symmetric positive definite.
//!
//! # Algorithm
//!
//! Uses Cholesky-based reduction to standard form:
//!
//! 1. Compute Cholesky decomposition: B = `LLᵀ`
//! 2. Transform: C = L⁻¹ A (L⁻¹)ᵀ (symmetric)
//! 3. Solve standard eigenproblem: Cy = λy
//! 4. Back-transform eigenvectors: x = (L⁻¹)ᵀ y
//!
//! # Applications
//!
//! - Hartree-Fock-Bogoliubov (HFB) nuclear structure
//! - Vibration analysis in structural mechanics
//! - Principal component analysis with constraints
//! - Quantum chemistry (Roothaan equations)
//!
//! # References
//!
//! - Golub & Van Loan, "Matrix Computations", Section 8.7
//! - Numerical Recipes, 3rd Edition, Section 11.0
//! - LAPACK DSYGV/DSYGVX routines

use crate::error::{BarracudaError, Result};
use crate::linalg::cholesky::{CholeskyDecomposition, cholesky_f64};
use crate::linalg::eigh::eigh_f64;
use std::sync::Arc;

use crate::device::WgpuDevice;

/// Generalized eigenvalue decomposition result
#[derive(Debug, Clone)]
pub struct GenEighDecomposition {
    /// Eigenvalues (λ) in ascending order
    pub eigenvalues: Vec<f64>,
    /// Eigenvectors (x) as column-major n×n matrix
    /// The i-th eigenvector is column i
    pub eigenvectors: Vec<f64>,
    /// Matrix dimension
    pub n: usize,
}

impl GenEighDecomposition {
    /// Get the i-th eigenvector
    /// Eigenvectors are stored as columns in row-major format:
    /// eigenvector j has component i at position [i * n + j]
    #[must_use]
    pub fn eigenvector(&self, i: usize) -> Option<Vec<f64>> {
        if i >= self.n {
            return None;
        }
        let mut v = vec![0.0; self.n];
        for (row, v_row) in v.iter_mut().enumerate() {
            *v_row = self.eigenvectors[row * self.n + i];
        }
        Some(v)
    }

    /// Sort eigenvalues and eigenvectors in descending order
    pub fn sort_descending(&mut self) {
        let n = self.n;
        let mut indices: Vec<usize> = (0..n).collect();
        indices.sort_by(|&a, &b| {
            self.eigenvalues[b]
                .partial_cmp(&self.eigenvalues[a])
                .unwrap_or(std::cmp::Ordering::Equal)
        });

        let old_eigenvalues = self.eigenvalues.clone();
        let old_eigenvectors = self.eigenvectors.clone();

        for (new_idx, &old_idx) in indices.iter().enumerate() {
            self.eigenvalues[new_idx] = old_eigenvalues[old_idx];
            // Copy column old_idx to column new_idx
            for row in 0..n {
                self.eigenvectors[row * n + new_idx] = old_eigenvectors[row * n + old_idx];
            }
        }
    }

    /// Verify that Ax = λBx for all eigenpairs
    /// Returns the maximum residual ||Ax - λBx|| / ||x||
    #[must_use]
    pub fn verify(&self, a: &[f64], b: &[f64]) -> f64 {
        let n = self.n;
        let mut max_residual: f64 = 0.0;

        for i in 0..n {
            let lambda = self.eigenvalues[i];

            // Extract eigenvector i (column i in row-major format)
            let x: Vec<f64> = (0..n).map(|row| self.eigenvectors[row * n + i]).collect();

            // Compute Ax
            let mut ax = vec![0.0; n];
            for row in 0..n {
                for col in 0..n {
                    ax[row] = a[row * n + col].mul_add(x[col], ax[row]);
                }
            }

            // Compute λBx
            let mut lbx = vec![0.0; n];
            for row in 0..n {
                for col in 0..n {
                    lbx[row] = (lambda * b[row * n + col]).mul_add(x[col], lbx[row]);
                }
            }

            // Compute ||Ax - λBx||
            let residual_norm: f64 = ax
                .iter()
                .zip(lbx.iter())
                .map(|(a, b)| (a - b).powi(2))
                .sum::<f64>()
                .sqrt();

            // Compute ||x||
            let x_norm: f64 = x.iter().map(|v| v.powi(2)).sum::<f64>().sqrt();

            let relative_residual = residual_norm / x_norm.max(1e-15);
            max_residual = max_residual.max(relative_residual);
        }

        max_residual
    }

    /// Compute the trace of the generalized eigenvalue problem
    /// (sum of eigenvalues)
    #[must_use]
    pub fn trace(&self) -> f64 {
        self.eigenvalues.iter().sum()
    }
}

/// Solve the generalized symmetric eigenvalue problem Ax = λBx
///
/// # Arguments
///
/// * `a` - Symmetric matrix A (row-major, n×n)
/// * `b` - Symmetric positive definite matrix B (row-major, n×n)
/// * `n` - Matrix dimension
///
/// # Returns
///
/// `GenEighDecomposition` with eigenvalues and eigenvectors
///
/// # Errors
///
/// - If B is not positive definite (Cholesky fails)
/// - If the transformed matrix has issues (eigh fails)
/// - If dimensions are invalid
///
/// # Example
///
/// ```no_run
/// use barracuda::linalg::gen_eigh_f64;
/// use barracuda::prelude::WgpuDevice;
/// use std::sync::Arc;
///
/// # async fn example() -> barracuda::error::Result<()> {
/// let device = Arc::new(WgpuDevice::new().await?);
/// let a = vec![2.0, 1.0, 1.0, 2.0];
/// let b = vec![1.0, 0.0, 0.0, 1.0];
///
/// let result = gen_eigh_f64(device, &a, &b, 2)?;
///
/// assert!((result.eigenvalues[0] - 1.0).abs() < 1e-10);
/// assert!((result.eigenvalues[1] - 3.0).abs() < 1e-10);
/// # Ok(())
/// # }
/// ```
pub fn gen_eigh_f64(
    device: Arc<WgpuDevice>,
    a: &[f64],
    b: &[f64],
    n: usize,
) -> Result<GenEighDecomposition> {
    if a.len() != n * n {
        return Err(BarracudaError::InvalidInput {
            message: format!(
                "Matrix A has wrong size: expected {}, got {}",
                n * n,
                a.len()
            ),
        });
    }
    if b.len() != n * n {
        return Err(BarracudaError::InvalidInput {
            message: format!(
                "Matrix B has wrong size: expected {}, got {}",
                n * n,
                b.len()
            ),
        });
    }

    // Step 1: Cholesky decomposition of B = LLᵀ (GPU)
    let chol = cholesky_f64(device, b, n)?;

    // Step 2: Compute C = L⁻¹ A (L⁻¹)ᵀ
    // First compute L⁻¹ A by solving L Y = A for each column of A
    let l_inv_a = solve_lower_triangular_matrix(&chol, a, n)?;

    // Then compute (L⁻¹ A) (L⁻¹)ᵀ = L⁻¹ A L⁻ᵀ
    // This is equivalent to solving L Z = (L⁻¹ A)ᵀ, then transposing
    let c = compute_symmetric_transform(&chol, &l_inv_a, n)?;

    // Step 3: Solve standard eigenproblem Cy = λy
    let std_eigh = eigh_f64(&c, n)?;

    // Step 4: Back-transform eigenvectors x = (L⁻¹)ᵀ y = L⁻ᵀ y
    // This is equivalent to solving Lᵀ x = y for each eigenvector
    let eigenvectors = back_transform_eigenvectors(&chol, &std_eigh.eigenvectors, n)?;

    Ok(GenEighDecomposition {
        eigenvalues: std_eigh.eigenvalues,
        eigenvectors,
        n,
    })
}

/// Solve LY = A where L is lower triangular (from Cholesky), column by column
fn solve_lower_triangular_matrix(
    chol: &CholeskyDecomposition,
    a: &[f64],
    n: usize,
) -> Result<Vec<f64>> {
    let l = &chol.l;
    let mut result = vec![0.0; n * n];

    // Solve for each column of A
    for col in 0..n {
        // Extract column of A
        let mut y = vec![0.0; n];
        for row in 0..n {
            y[row] = a[row * n + col];
        }

        // Forward substitution: Ly = a_col
        for i in 0..n {
            let mut sum = y[i];
            for j in 0..i {
                sum = l[i * n + j].mul_add(-y[j], sum);
            }
            y[i] = sum / l[i * n + i];
        }

        // Store result column
        for row in 0..n {
            result[row * n + col] = y[row];
        }
    }

    Ok(result)
}

/// Compute C = (L⁻¹ A) (L⁻¹)ᵀ = Y Yᵀ where Y = L⁻¹ A
/// But we want C = L⁻¹ A L⁻ᵀ, so we need to solve Lᵀ Z = Yᵀ row by row
fn compute_symmetric_transform(
    chol: &CholeskyDecomposition,
    l_inv_a: &[f64],
    n: usize,
) -> Result<Vec<f64>> {
    let l = &chol.l;
    let mut c = vec![0.0; n * n];

    // For each row of C, solve Lᵀ z = (L⁻¹ A)[row,:]ᵀ
    // Then C[row,:] = zᵀ
    // But since C should be symmetric, we can compute C = (L⁻¹ A) (L⁻ᵀ)
    // which is (L⁻¹ A) (L⁻¹)ᵀ

    // Actually, let's compute it directly:
    // C[i,j] = sum_k (L⁻¹ A)[i,k] * (L⁻ᵀ)[k,j]
    //        = sum_k Y[i,k] * (L⁻¹)[j,k]

    // Compute L⁻¹ (inverse of lower triangular)
    let mut l_inv = vec![0.0; n * n];
    for i in 0..n {
        // Solve L x = e_i
        let mut x = vec![0.0; n];
        x[i] = 1.0;

        for row in 0..n {
            let mut sum = if row == i { 1.0 } else { 0.0 };
            for k in 0..row {
                sum = l[row * n + k].mul_add(-x[k], sum);
            }
            x[row] = sum / l[row * n + row];
        }

        // Store as column i of L⁻¹
        for row in 0..n {
            l_inv[row * n + i] = x[row];
        }
    }

    // C = Y * L⁻ᵀ = (L⁻¹ A) * (L⁻¹)ᵀ
    for i in 0..n {
        for j in 0..n {
            let mut sum = 0.0;
            for k in 0..n {
                // Y[i,k] * L⁻ᵀ[k,j] = Y[i,k] * L⁻¹[j,k]
                sum = l_inv_a[i * n + k].mul_add(l_inv[j * n + k], sum);
            }
            c[i * n + j] = sum;
        }
    }

    // Ensure symmetry (average out numerical errors)
    for i in 0..n {
        for j in i + 1..n {
            let avg = 0.5 * (c[i * n + j] + c[j * n + i]);
            c[i * n + j] = avg;
            c[j * n + i] = avg;
        }
    }

    Ok(c)
}

/// Back-transform eigenvectors: x = L⁻ᵀ y (solve Lᵀ x = y)
///
/// Note: eigenvectors from `eigh_f64` are stored as columns in row-major format:
/// eigenvector j has component i at position [i * n + j]
fn back_transform_eigenvectors(
    chol: &CholeskyDecomposition,
    y: &[f64],
    n: usize,
) -> Result<Vec<f64>> {
    let l = &chol.l;
    let mut x = vec![0.0; n * n];

    // For each eigenvector (column j)
    for j in 0..n {
        // Extract eigenvector j: y_j[i] = y[i * n + j]
        let mut xi = vec![0.0; n];
        for i in 0..n {
            xi[i] = y[i * n + j];
        }

        // Solve Lᵀ x = y (backward substitution)
        // Lᵀ is upper triangular, L is lower triangular stored row-major
        // Lᵀ[i,k] = L[k,i] = l[k * n + i]
        for i in (0..n).rev() {
            let mut sum = xi[i];
            for k in i + 1..n {
                // Lᵀ[i,k] * xi[k] = L[k,i] * xi[k]
                sum = l[k * n + i].mul_add(-xi[k], sum);
            }
            xi[i] = sum / l[i * n + i];
        }

        // Store result in same format: x_j[i] = x[i * n + j]
        for i in 0..n {
            x[i * n + j] = xi[i];
        }
    }

    Ok(x)
}

/// Solve generalized eigenvalue problem with B = I (identity)
///
/// This is a convenience function that reduces to standard eigenvalue problem.
///
/// # Errors
///
/// Returns [`Err`] if matrix dimensions are invalid or eigendecomposition fails.
pub fn gen_eigh_identity_b(a: &[f64], n: usize) -> Result<GenEighDecomposition> {
    let eigh = eigh_f64(a, n)?;

    Ok(GenEighDecomposition {
        eigenvalues: eigh.eigenvalues,
        eigenvectors: eigh.eigenvectors,
        n,
    })
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::device::test_pool::get_test_device_if_f64_gpu_available_sync;

    #[test]
    fn test_gen_eigh_identity_b() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        // When B = I, gen_eigh should give same results as standard eigh
        let a = vec![4.0, 1.0, 1.0, 3.0];
        let b = vec![1.0, 0.0, 0.0, 1.0]; // Identity

        let result = gen_eigh_f64(device, &a, &b, 2).unwrap();

        // Standard eigenvalues of [[4, 1], [1, 3]]
        // Characteristic polynomial: (4-λ)(3-λ) - 1 = λ² - 7λ + 11
        // λ = (7 ± √5) / 2 ≈ 2.382, 4.618
        let expected_min = (7.0 - 5.0_f64.sqrt()) / 2.0;
        let expected_max = f64::midpoint(7.0, 5.0_f64.sqrt());

        assert!(
            (result.eigenvalues[0] - expected_min).abs() < 1e-10,
            "Expected λ₀ ≈ {}, got {}",
            expected_min,
            result.eigenvalues[0]
        );
        assert!(
            (result.eigenvalues[1] - expected_max).abs() < 1e-10,
            "Expected λ₁ ≈ {}, got {}",
            expected_max,
            result.eigenvalues[1]
        );
    }

    #[test]
    fn test_gen_eigh_scaled_b() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        // If B = 2I, eigenvalues should be halved
        let a = vec![4.0, 0.0, 0.0, 6.0];
        let b = vec![2.0, 0.0, 0.0, 2.0]; // 2I

        let result = gen_eigh_f64(device, &a, &b, 2).unwrap();

        // A has eigenvalues 4, 6
        // Gen. eigenvalues with B = 2I should be 4/2 = 2, 6/2 = 3
        assert!(
            (result.eigenvalues[0] - 2.0).abs() < 1e-10,
            "Expected λ₀ = 2, got {}",
            result.eigenvalues[0]
        );
        assert!(
            (result.eigenvalues[1] - 3.0).abs() < 1e-10,
            "Expected λ₁ = 3, got {}",
            result.eigenvalues[1]
        );
    }

    #[test]
    fn test_gen_eigh_verify() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        // Test that Ax = λBx holds
        let a = vec![5.0, 2.0, 2.0, 3.0];
        let b = vec![2.0, 1.0, 1.0, 2.0]; // SPD

        let result = gen_eigh_f64(device, &a, &b, 2).unwrap();

        let residual = result.verify(&a, &b);
        assert!(residual < 1e-10, "Residual too large: {residual}");
    }

    #[test]
    fn test_gen_eigh_3x3() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        // 3×3 example
        #[rustfmt::skip]
        let a = vec![
            6.0, 2.0, 1.0,
            2.0, 5.0, 2.0,
            1.0, 2.0, 4.0,
        ];
        #[rustfmt::skip]
        let b = vec![
            2.0, 0.5, 0.0,
            0.5, 2.0, 0.5,
            0.0, 0.5, 2.0,
        ];

        let result = gen_eigh_f64(device, &a, &b, 3).unwrap();

        // Verify all eigenpairs
        let residual = result.verify(&a, &b);
        assert!(residual < 1e-9, "Residual too large: {residual}");
    }

    #[test]
    fn test_gen_eigh_eigenvector_access() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        let a = vec![2.0, 1.0, 1.0, 2.0];
        let b = vec![1.0, 0.0, 0.0, 1.0];

        let result = gen_eigh_f64(device, &a, &b, 2).unwrap();

        let v0 = result.eigenvector(0).unwrap();
        let v1 = result.eigenvector(1).unwrap();

        assert_eq!(v0.len(), 2);
        assert_eq!(v1.len(), 2);

        // Out of bounds
        assert!(result.eigenvector(2).is_none());
    }

    #[test]
    fn test_gen_eigh_sort_descending() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        let a = vec![1.0, 0.0, 0.0, 3.0];
        let b = vec![1.0, 0.0, 0.0, 1.0];

        let mut result = gen_eigh_f64(device, &a, &b, 2).unwrap();

        // Initially ascending (1, 3)
        assert!(result.eigenvalues[0] < result.eigenvalues[1]);

        result.sort_descending();

        // After sorting: descending (3, 1)
        assert!(result.eigenvalues[0] > result.eigenvalues[1]);
        assert!((result.eigenvalues[0] - 3.0).abs() < 1e-10);
        assert!((result.eigenvalues[1] - 1.0).abs() < 1e-10);
    }

    #[test]
    fn test_gen_eigh_non_spd_b_error() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        let a = vec![1.0, 0.0, 0.0, 1.0];
        // B is not positive definite (has negative eigenvalue)
        let b = vec![1.0, 2.0, 2.0, 1.0]; // Eigenvalues: 3, -1

        let result = gen_eigh_f64(device, &a, &b, 2);
        assert!(result.is_err());
    }

    #[test]
    fn test_gen_eigh_dimension_mismatch() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        let a = vec![1.0, 0.0, 0.0, 1.0];
        let b = vec![1.0, 0.0, 0.0]; // Wrong size

        let result = gen_eigh_f64(device, &a, &b, 2);
        assert!(result.is_err());
    }

    #[test]
    fn test_gen_eigh_trace() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        let a = vec![3.0, 1.0, 1.0, 3.0];
        let b = vec![1.0, 0.0, 0.0, 1.0];
        let result = gen_eigh_f64(device, &a, &b, 2).unwrap();

        // With B = I, trace of eigenvalues = trace of A = 6
        let trace = result.trace();
        assert!((trace - 6.0).abs() < 1e-10, "Expected trace 6, got {trace}");
    }

    #[test]
    fn test_gen_eigh_larger_matrix() {
        let Some(device) = get_test_device_if_f64_gpu_available_sync() else {
            return;
        };
        let n = 5;
        let mut a = vec![0.0; n * n];
        let mut b = vec![0.0; n * n];
        for i in 0..n {
            for j in 0..n {
                a[i * n + j] = if i == j { 5.0 } else { 0.5 };
                b[i * n + j] = if i == j { 2.0 } else { 0.1 };
            }
        }
        let result = gen_eigh_f64(device, &a, &b, n).unwrap();

        let residual = result.verify(&a, &b);
        assert!(residual < 1e-8, "Residual too large for 5×5: {residual}");
    }
}