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
// SPDX-License-Identifier: AGPL-3.0-or-later
//! Observable Computation for Molecular Dynamics — Shader-First Architecture.
//!
//! All underlying math originates as f64 WGSL shaders. High-level analysis
//! functions (`compute_rdf`, `compute_vacf`, etc.) orchestrate snapshot
//! iteration and normalization around GPU kernels.
//!
//! **GPU Observables** (shader-dispatched):
//! - [`KineticEnergyF64`] — per-particle KE via `kinetic_energy_f64.wgsl`
//! - [`RdfHistogramF64`] — pair-distance histogram via `rdf_histogram_f64.wgsl`
//! - [`SsfGpu`] — static structure factor via `ssf_f64.wgsl`
//! - [`VacfGpu`] / [`VacfBatchGpu`] — velocity autocorrelation via `vacf_f64.wgsl`
//! - [`MsdGpu`] — mean-squared displacement via `msd_f64.wgsl`
//! - [`HeatCurrentGpu`] — heat current via `heat_current_f64.wgsl`
//! - [`StressVirialGpu`] — stress tensor via `stress_virial_f64.wgsl`
//!
//! **Snapshot orchestration** (control flow only, math in shaders):
//! - `compute_rdf` / `compute_vacf` / `compute_ssf` / `compute_msd`
//! - `validate_energy` — CPU statistics over scalar energy history
//!
//! **Deep Debt Compliance**:
//! - ✅ WGSL shader-first (all math as .wgsl, CPU gated `#[cfg(test)]`)
//! - ✅ Full f64 precision
//! - ✅ Zero unsafe code

pub mod heat_current_gpu;
mod kinetic_energy;
mod kinetic_energy_f64;
pub mod msd_gpu;
mod rdf_f64;
mod ssf_gpu;
pub mod transport_gpu;
mod vacf_gpu;

pub use heat_current_gpu::HeatCurrentGpu;
pub use kinetic_energy::KineticEnergy;
pub use kinetic_energy_f64::KineticEnergyF64;
pub use msd_gpu::MsdGpu;
pub use rdf_f64::RdfHistogramF64;
pub use ssf_gpu::SsfGpu;
pub use transport_gpu::{GpuVelocityRing, StressVirialGpu, VacfBatchGpu};
pub use vacf_gpu::VacfGpu;

use std::f64::consts::PI;

/// RDF result: g(r) binned at discrete r values
#[derive(Clone, Debug)]
pub struct Rdf {
    /// Radial distances (bin centers)
    pub r_values: Vec<f64>,
    /// Pair correlation function g(r) at each r
    pub g_values: Vec<f64>,
    /// Bin width
    pub dr: f64,
}

/// VACF result: C(t) at discrete lag times
#[derive(Clone, Debug)]
pub struct Vacf {
    /// Lag times
    pub t_values: Vec<f64>,
    /// Normalized autocorrelation C(t)/C(0)
    pub c_values: Vec<f64>,
    /// Diffusion coefficient from Green-Kubo integral
    pub diffusion_coeff: f64,
}

/// Energy validation result
#[derive(Clone, Debug)]
pub struct EnergyValidation {
    /// Mean total energy (after skip)
    pub mean_total: f64,
    /// Standard deviation of total energy
    pub std_total: f64,
    /// Drift percentage from initial to final
    pub drift_pct: f64,
    /// Mean temperature (if computed)
    pub mean_temperature: f64,
    /// Standard deviation of temperature
    pub std_temperature: f64,
    /// Whether drift is within acceptable threshold
    pub passed: bool,
}

/// Compute RDF from position snapshots (CPU post-process)
///
/// # Arguments
/// * `snapshots` - Position snapshots, each [N*3] flattened
/// * `n` - Number of particles
/// * `box_side` - Box side length in reduced units
/// * `n_bins` - Number of histogram bins
#[must_use]
pub fn compute_rdf(snapshots: &[Vec<f64>], n: usize, box_side: f64, n_bins: usize) -> Rdf {
    let r_max = box_side / 2.0;
    let dr = r_max / n_bins as f64;
    let mut histogram = vec![0u64; n_bins];
    let n_frames = snapshots.len();

    for snap in snapshots {
        for i in 0..n {
            let xi = snap[i * 3];
            let yi = snap[i * 3 + 1];
            let zi = snap[i * 3 + 2];

            for j in (i + 1)..n {
                let mut dx = snap[j * 3] - xi;
                let mut dy = snap[j * 3 + 1] - yi;
                let mut dz = snap[j * 3 + 2] - zi;

                // PBC minimum image
                dx = box_side.mul_add(-(dx / box_side).round(), dx);
                dy = box_side.mul_add(-(dy / box_side).round(), dy);
                dz = box_side.mul_add(-(dz / box_side).round(), dz);

                let r = dz.mul_add(dz, dx.mul_add(dx, dy * dy)).sqrt();
                let bin = (r / dr) as usize;
                if bin < n_bins {
                    histogram[bin] += 1;
                }
            }
        }
    }

    // Normalize: g(r) = histogram / (n_frames * N * n_density * 4π r² dr)
    let n_density = 3.0 / (4.0 * PI); // OCP reduced units
    let n_f = n as f64;
    let r_values: Vec<f64> = (0..n_bins).map(|i| (i as f64 + 0.5) * dr).collect();
    let g_values: Vec<f64> = r_values
        .iter()
        .enumerate()
        .map(|(i, &r)| {
            let shell_vol = 4.0 * PI * r * r * dr;
            2.0 * histogram[i] as f64 / (n_frames as f64 * n_f * n_density * shell_vol).max(1e-30)
        })
        .collect();

    Rdf {
        r_values,
        g_values,
        dr,
    }
}

/// Compute VACF from velocity snapshots (CPU post-process)
///
/// # Arguments
/// * `vel_snapshots` - Velocity snapshots, each [N*3] flattened
/// * `n` - Number of particles
/// * `dt_dump` - Time between snapshots (reduced units)
/// * `max_lag` - Maximum lag in snapshots
#[must_use]
pub fn compute_vacf(vel_snapshots: &[Vec<f64>], n: usize, dt_dump: f64, max_lag: usize) -> Vacf {
    let n_frames = vel_snapshots.len();
    let n_lag = max_lag.min(n_frames);
    let mut c_values = vec![0.0f64; n_lag];
    let mut counts = vec![0usize; n_lag];

    for t0 in 0..n_frames {
        for lag in 0..n_lag {
            let t1 = t0 + lag;
            if t1 >= n_frames {
                break;
            }
            let v0 = &vel_snapshots[t0];
            let v1 = &vel_snapshots[t1];

            let mut dot_sum = 0.0;
            for i in 0..n {
                dot_sum += v0[i * 3 + 2].mul_add(
                    v1[i * 3 + 2],
                    v0[i * 3].mul_add(v1[i * 3], v0[i * 3 + 1] * v1[i * 3 + 1]),
                );
            }
            c_values[lag] += dot_sum / n as f64;
            counts[lag] += 1;
        }
    }

    // Average over time origins
    for i in 0..n_lag {
        if counts[i] > 0 {
            c_values[i] /= counts[i] as f64;
        }
    }

    // Normalize by C(0)
    let c0 = c_values[0].max(1e-30);
    let c_normalized: Vec<f64> = c_values.iter().map(|&c| c / c0).collect();

    // Diffusion coefficient: D* = (1/3) integral_0^inf C(t) dt
    let mut integral = 0.0;
    for i in 1..n_lag {
        integral = (0.5 * (c_values[i - 1] + c_values[i])).mul_add(dt_dump, integral);
    }
    let diffusion_coeff = integral / 3.0;

    let t_values: Vec<f64> = (0..n_lag).map(|i| i as f64 * dt_dump).collect();

    Vacf {
        t_values,
        c_values: c_normalized,
        diffusion_coeff,
    }
}

/// Compute static structure factor S(k) from position snapshots
///
/// # Arguments
/// * `snapshots` - Position snapshots, each [N*3] flattened
/// * `n` - Number of particles
/// * `box_side` - Box side length in reduced units
/// * `max_k_harmonics` - Number of k-vectors along each axis
#[must_use]
pub fn compute_ssf(
    snapshots: &[Vec<f64>],
    n: usize,
    box_side: f64,
    max_k_harmonics: usize,
) -> Vec<(f64, f64)> {
    let dk = 2.0 * PI / box_side;
    let _n_frames = snapshots.len(); // Used implicitly via iteration
    let mut sk_values: Vec<(f64, f64)> = Vec::new();

    for kn in 1..=max_k_harmonics {
        let k_mag = kn as f64 * dk;
        let mut sk_sum = 0.0;
        let mut count = 0;

        for snap in snapshots {
            // S(k) = <|rho(k)|²> / N along each principal axis
            for axis in 0..3 {
                let mut re = 0.0;
                let mut im = 0.0;
                for j in 0..n {
                    let r_component = snap[j * 3 + axis];
                    let phase = k_mag * r_component;
                    re += phase.cos();
                    im += phase.sin();
                }
                sk_sum += (re * re + im * im) / n as f64;
                count += 1;
            }
        }

        sk_values.push((k_mag, sk_sum / count as f64));
    }

    sk_values
}

/// Mean-Squared Displacement (MSD) result.
#[derive(Clone, Debug)]
pub struct Msd {
    /// Lag times τ in reduced units
    pub t_values: Vec<f64>,
    /// MSD(τ) = <|r(t+τ) - r(t)|²>
    pub msd_values: Vec<f64>,
    /// Diffusion coefficient from Einstein relation: D* = lim_{τ→∞} MSD(τ)/(6τ)
    pub diffusion_coeff: f64,
}

/// Compute Mean-Squared Displacement from position snapshots
///
/// MSD is fundamental for diffusion analysis:
/// - Einstein relation: D* = lim_{τ→∞} MSD(τ) / (6τ)
/// - Slope of MSD vs τ plot indicates diffusion regime
/// - Linear → normal diffusion, sub-linear → sub-diffusion
///
/// # Arguments
/// * `snapshots` - Position snapshots, each [N*3] flattened
/// * `n` - Number of particles
/// * `box_side` - Box side length (for PBC unwrapping)
/// * `dt` - Time between snapshots
/// * `max_lag` - Maximum lag (number of snapshot intervals)
#[must_use]
pub fn compute_msd(
    snapshots: &[Vec<f64>],
    n: usize,
    box_side: f64,
    dt: f64,
    max_lag: usize,
) -> Msd {
    let n_frames = snapshots.len();
    let actual_max_lag = max_lag.min(n_frames - 1);

    let mut t_values = Vec::with_capacity(actual_max_lag);
    let mut msd_values = Vec::with_capacity(actual_max_lag);

    // Unwrap positions (handle PBC jumps)
    let unwrapped = unwrap_positions(snapshots, n, box_side);

    for lag in 1..=actual_max_lag {
        let mut msd_sum = 0.0;
        let mut count = 0;

        for t0 in 0..(n_frames - lag) {
            let t1 = t0 + lag;

            for i in 0..n {
                let dx = unwrapped[t1][i * 3] - unwrapped[t0][i * 3];
                let dy = unwrapped[t1][i * 3 + 1] - unwrapped[t0][i * 3 + 1];
                let dz = unwrapped[t1][i * 3 + 2] - unwrapped[t0][i * 3 + 2];
                msd_sum += dz.mul_add(dz, dx.mul_add(dx, dy * dy));
                count += 1;
            }
        }

        let msd = msd_sum / count as f64;
        t_values.push(lag as f64 * dt);
        msd_values.push(msd);
    }

    // Estimate diffusion coefficient from long-time slope
    // D* = MSD / (6 * t) — use last quarter of data for fitting
    let fit_start = (actual_max_lag * 3 / 4).max(1);
    let diffusion_coeff = if fit_start < actual_max_lag {
        let t_fit: Vec<f64> = t_values[fit_start..].to_vec();
        let msd_fit: Vec<f64> = msd_values[fit_start..].to_vec();
        linear_fit_slope(&t_fit, &msd_fit) / 6.0
    } else if let (Some(&msd), Some(&t)) = (msd_values.last(), t_values.last()) {
        msd / (6.0 * t)
    } else {
        0.0
    };

    Msd {
        t_values,
        msd_values,
        diffusion_coeff,
    }
}

/// Unwrap positions to handle PBC jumps
///
/// Detects when particles cross box boundaries and adjusts coordinates
/// to produce continuous trajectories for MSD/VACF calculations.
fn unwrap_positions(snapshots: &[Vec<f64>], n: usize, box_side: f64) -> Vec<Vec<f64>> {
    if snapshots.is_empty() {
        return Vec::new();
    }

    let mut unwrapped = Vec::with_capacity(snapshots.len());
    unwrapped.push(snapshots[0].clone());

    let mut image_counts = vec![[0i32; 3]; n]; // Track box crossings

    for frame_idx in 1..snapshots.len() {
        let prev = &unwrapped[frame_idx - 1];
        let curr = &snapshots[frame_idx];
        let mut unwrapped_frame = Vec::with_capacity(n * 3);

        for i in 0..n {
            for d in 0..3 {
                let r_prev = prev[i * 3 + d];
                let r_curr = curr[i * 3 + d];

                // Detect box crossing
                let delta = r_curr - (image_counts[i][d] as f64).mul_add(-box_side, r_prev);
                if delta > box_side / 2.0 {
                    image_counts[i][d] -= 1;
                } else if delta < -box_side / 2.0 {
                    image_counts[i][d] += 1;
                }

                unwrapped_frame.push((image_counts[i][d] as f64).mul_add(box_side, r_curr));
            }
        }

        unwrapped.push(unwrapped_frame);
    }

    unwrapped
}

/// Simple linear fit to get slope (for diffusion coefficient)
fn linear_fit_slope(x: &[f64], y: &[f64]) -> f64 {
    if x.len() < 2 {
        return 0.0;
    }

    let n = x.len() as f64;
    let sum_x: f64 = x.iter().sum();
    let sum_y: f64 = y.iter().sum();
    let sum_xy: f64 = x.iter().zip(y.iter()).map(|(xi, yi)| xi * yi).sum();
    let sum_x2: f64 = x.iter().map(|xi| xi * xi).sum();

    let denom = n.mul_add(sum_x2, -(sum_x * sum_x));
    if denom.abs() < 1e-30 {
        return 0.0;
    }

    n.mul_add(sum_xy, -(sum_x * sum_y)) / denom
}

/// Validate energy conservation from energy history
///
/// # Arguments
/// * `energies` - Vector of (step, KE, PE, total) tuples
/// * `skip_fraction` - Fraction of initial data to skip (e.g., 0.1)
#[must_use]
pub fn validate_energy(
    energies: &[(usize, f64, f64, f64)],
    skip_fraction: f64,
) -> EnergyValidation {
    if energies.is_empty() {
        return EnergyValidation {
            mean_total: 0.0,
            std_total: 0.0,
            drift_pct: 0.0,
            mean_temperature: 0.0,
            std_temperature: 0.0,
            passed: false,
        };
    }

    let skip = ((energies.len() as f64) * skip_fraction) as usize;
    let stable: Vec<_> = energies.iter().skip(skip).collect();

    if stable.is_empty() {
        return EnergyValidation {
            mean_total: 0.0,
            std_total: 0.0,
            drift_pct: 0.0,
            mean_temperature: 0.0,
            std_temperature: 0.0,
            passed: false,
        };
    }

    let totals: Vec<f64> = stable.iter().map(|e| e.3).collect();
    let mean_e: f64 = totals.iter().sum::<f64>() / totals.len() as f64;
    let var_e: f64 = totals.iter().map(|e| (e - mean_e).powi(2)).sum::<f64>() / totals.len() as f64;
    let std_e = var_e.sqrt();

    let e_initial = stable[0].3;
    let e_final = stable[stable.len() - 1].3;
    let drift_pct = if mean_e.abs() > 1e-30 {
        ((e_final - e_initial) / mean_e.abs()).abs() * 100.0
    } else {
        0.0
    };

    let passed = drift_pct < 5.0;

    EnergyValidation {
        mean_total: mean_e,
        std_total: std_e,
        drift_pct,
        mean_temperature: 0.0, // Caller should compute from KE
        std_temperature: 0.0,
        passed,
    }
}

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

    #[test]
    fn test_msd_ballistic_motion() {
        // Test MSD with ballistic (constant velocity) motion
        // MSD should grow as t² for ballistic regime
        let n = 2;
        let box_side = 100.0; // Large box, no PBC effects
        let dt = 0.1;
        let velocity = 1.0;

        // Create trajectory: particles moving with constant velocity
        let n_frames = 50;
        let mut snapshots = Vec::with_capacity(n_frames);

        for frame in 0..n_frames {
            let t = frame as f64 * dt;
            snapshots.push(vec![
                t * velocity,  // particle 0: x
                0.0,           // particle 0: y
                0.0,           // particle 0: z
                -t * velocity, // particle 1: x (opposite direction)
                0.0,           // particle 1: y
                0.0,           // particle 1: z
            ]);
        }

        let msd = compute_msd(&snapshots, n, box_side, dt, 20);

        // For ballistic motion: MSD(t) = v² * t² * 2 (average over two particles)
        // At lag=10 (t=1.0): expected MSD ≈ 1² * 1.0² * 2 = 2.0
        assert!(!msd.msd_values.is_empty());
        assert_eq!(msd.msd_values.len(), 20);

        // Check quadratic growth (MSD ∝ t² for ballistic)
        let msd_early = msd.msd_values[4]; // lag=5
        let msd_late = msd.msd_values[9]; // lag=10
        let ratio = msd_late / msd_early;
        // Should be close to (10/5)² = 4
        assert!((ratio - 4.0).abs() < 0.1, "Ballistic MSD ratio: {ratio}");
    }

    #[test]
    fn test_msd_diffusive_motion() {
        // Test MSD with random walk (diffusive motion)
        // MSD should grow linearly as 6*D*t for diffusive regime
        let n = 1;
        let box_side = 1000.0;
        let dt = 1.0;

        // Create random walk trajectory
        let n_frames = 100;
        let mut snapshots = Vec::with_capacity(n_frames);
        let step_size = 0.5;

        // Deterministic "random" walk for reproducibility
        let mut x = 0.0;
        let mut y = 0.0;
        let mut z = 0.0;

        // Pseudo-random multipliers (deliberately close to golden ratio and e)
        #[expect(clippy::approx_constant, reason = "physics constant")]
        let (mult_phi, mult_e) = (1.618, 2.718);
        for frame in 0..n_frames {
            snapshots.push(vec![x, y, z]);

            // Pseudo-random step based on frame number
            let angle1 = (frame as f64 * mult_phi) % (2.0 * PI);
            let angle2 = (frame as f64 * mult_e) % PI;
            x = (step_size * angle2.sin()).mul_add(angle1.cos(), x);
            y = (step_size * angle2.sin()).mul_add(angle1.sin(), y);
            z += step_size * angle2.cos();
        }

        let msd = compute_msd(&snapshots, n, box_side, dt, 50);

        // MSD should be positive and generally increasing
        assert!(!msd.msd_values.is_empty());
        assert!(msd.msd_values[0] > 0.0);
        assert!(msd.diffusion_coeff >= 0.0);
    }

    #[test]
    fn test_msd_pbc_unwrapping() {
        // Test that PBC unwrapping works correctly
        let n = 1;
        let box_side = 10.0;
        let dt = 1.0;

        // Particle crosses box boundary
        let snapshots = vec![
            vec![9.0, 0.0, 0.0], // Near right edge
            vec![1.0, 0.0, 0.0], // Wrapped to left (actual displacement +2)
            vec![3.0, 0.0, 0.0], // Continues right
        ];

        let msd = compute_msd(&snapshots, n, box_side, dt, 2);

        // After unwrapping: 9 → 11 → 13 (not 9 → 1 → 3)
        // MSD at lag=1 should be ~4 (displacement of 2)
        // MSD at lag=2 should be ~16 (displacement of 4)
        assert!(msd.msd_values.len() == 2);
        assert!(
            (msd.msd_values[0] - 4.0).abs() < 0.5,
            "MSD[0] = {}",
            msd.msd_values[0]
        );
        assert!(
            (msd.msd_values[1] - 16.0).abs() < 0.5,
            "MSD[1] = {}",
            msd.msd_values[1]
        );
    }

    #[test]
    fn test_linear_fit_slope() {
        // Test linear fitting
        let x = vec![1.0, 2.0, 3.0, 4.0, 5.0];
        let y = vec![2.0, 4.0, 6.0, 8.0, 10.0]; // y = 2x

        let slope = linear_fit_slope(&x, &y);
        assert!((slope - 2.0).abs() < 1e-10);
    }
}