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
// SPDX-License-Identifier: AGPL-3.0-or-later

//! Biosignal processing — EDA, stress detection, beat classification.
//!
//! Absorbed from healthSpring V19 (Exp081–082, Exp084).

// ── Signal Processing Primitives ─────────────────────────────────────────────

/// Efficient O(n) rolling average using a running sum accumulator.
///
/// For each output sample, computes the mean of the surrounding `window_size`
/// input samples (centered). 21x faster than naive convolution for large windows
/// (healthSpring P1 optimization request).
#[must_use]
pub fn rolling_average(signal: &[f64], window_size: usize) -> Vec<f64> {
    if signal.is_empty() || window_size == 0 {
        return vec![];
    }
    let n = signal.len();
    let ws = window_size.min(n);
    let half = ws / 2;
    let mut result = Vec::with_capacity(n);
    let mut sum: f64 = signal[..ws.min(n)].iter().sum();
    let mut left = 0usize;
    let mut right = ws.min(n);

    for i in 0..n {
        let target_left = i.saturating_sub(half);
        let target_right = (i + half + 1).min(n);

        while left < target_left {
            sum -= signal[left];
            left += 1;
        }
        while right < target_right {
            sum += signal[right];
            right += 1;
        }
        while left > target_left && left > 0 {
            left -= 1;
            sum += signal[left];
        }

        let count = right - left;
        result.push(if count > 0 { sum / count as f64 } else { 0.0 });
    }
    result
}

/// 1D convolution of signal with kernel (valid output).
///
/// Returns a vector of length `signal.len() - kernel.len() + 1`.
#[must_use]
pub fn convolve_1d(signal: &[f64], kernel: &[f64]) -> Vec<f64> {
    if kernel.is_empty() || signal.len() < kernel.len() {
        return vec![];
    }

    #[cfg(feature = "cpu-shader")]
    {
        if let Ok(out) = convolve_1d_shader(signal, kernel) {
            return out;
        }
    }

    convolve_1d_scalar(signal, kernel)
}

fn convolve_1d_scalar(signal: &[f64], kernel: &[f64]) -> Vec<f64> {
    let out_len = signal.len() - kernel.len() + 1;
    (0..out_len)
        .map(|i| {
            signal[i..i + kernel.len()]
                .iter()
                .zip(kernel.iter())
                .map(|(&s, &k)| s * k)
                .sum()
        })
        .collect()
}

#[cfg(feature = "cpu-shader")]
fn convolve_1d_shader(signal: &[f64], kernel: &[f64]) -> crate::error::Result<Vec<f64>> {
    use crate::unified_hardware::{CpuShaderDispatch, ShaderBinding, ShaderDispatch};

    let wgsl = include_str!("../shaders/health/convolve_1d_f64.wgsl");
    let dispatcher = CpuShaderDispatch::new();
    let out_len = signal.len() - kernel.len() + 1;

    let mut sig_buf: Vec<u8> = signal.iter().flat_map(|v| v.to_le_bytes()).collect();
    let mut kern_buf: Vec<u8> = kernel.iter().flat_map(|v| v.to_le_bytes()).collect();
    let mut out_buf = vec![0u8; out_len * 8];

    // Params { signal_len: u32, kernel_len: u32 } = 8 bytes
    let mut params_buf = vec![0u8; 8];
    params_buf[..4].copy_from_slice(&(signal.len() as u32).to_le_bytes());
    params_buf[4..8].copy_from_slice(&(kernel.len() as u32).to_le_bytes());

    let mut bindings = vec![
        ShaderBinding {
            group: 0,
            binding: 0,
            data: &mut sig_buf,
            read_only: true,
        },
        ShaderBinding {
            group: 0,
            binding: 1,
            data: &mut kern_buf,
            read_only: true,
        },
        ShaderBinding {
            group: 0,
            binding: 2,
            data: &mut out_buf,
            read_only: false,
        },
        ShaderBinding {
            group: 0,
            binding: 3,
            data: &mut params_buf,
            read_only: true,
        },
    ];

    let workgroups = (
        (out_len as u32).div_ceil(crate::device::capabilities::WORKGROUP_SIZE_1D),
        1,
        1,
    );
    dispatcher.dispatch_wgsl(wgsl, "main", &mut bindings, workgroups)?;

    Ok(out_buf
        .chunks_exact(8)
        .map(|c| f64::from_le_bytes([c[0], c[1], c[2], c[3], c[4], c[5], c[6], c[7]]))
        .collect())
}

// ── EDA (Electrodermal Activity) ─────────────────────────────────────────────

/// Tonic skin conductance level via moving-window integration.
#[must_use]
pub fn eda_scl(signal: &[f64], window_size: usize) -> Vec<f64> {
    if signal.is_empty() || window_size == 0 {
        return vec![];
    }
    let ws = window_size.min(signal.len());
    let mut result = Vec::with_capacity(signal.len());
    let mut sum: f64 = signal[..ws].iter().sum();
    result.push(sum / ws as f64);

    for i in 1..signal.len() {
        if i + ws - 1 < signal.len() {
            sum += signal[i + ws - 1];
        }
        if i >= 1 && i - 1 + ws <= signal.len() {
            sum -= signal[i - 1];
        }
        let count = ws.min(signal.len() - i);
        result.push(sum / count as f64);
    }
    result
}

/// Phasic SCR component: signal minus tonic SCL.
#[must_use]
pub fn eda_phasic(signal: &[f64], window_size: usize) -> Vec<f64> {
    let scl = eda_scl(signal, window_size);
    signal
        .iter()
        .zip(scl.iter())
        .map(|(&s, &t)| (s - t).max(0.0))
        .collect()
}

/// Detect SCR peaks in the phasic component.
///
/// Returns indices where the phasic component exceeds `threshold`.
#[must_use]
pub fn eda_detect_scr(phasic: &[f64], threshold: f64) -> Vec<usize> {
    let mut peaks = Vec::new();
    for i in 1..phasic.len().saturating_sub(1) {
        if phasic[i] > threshold && phasic[i] > phasic[i - 1] && phasic[i] >= phasic[i + 1] {
            peaks.push(i);
        }
    }
    peaks
}

// ── Stress Detection ─────────────────────────────────────────────────────────

/// Stress assessment result.
#[derive(Debug, Clone)]
pub struct StressAssessment {
    /// Composite stress index in `[0, 1]`.
    pub stress_index: f64,
    /// SCR events per minute.
    pub scr_rate: f64,
    /// Categorical stress level.
    pub level: StressLevel,
}

/// Categorical stress level.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum StressLevel {
    /// Low stress (`< 0.25`).
    Low,
    /// Moderate stress (`0.25..0.50`).
    Moderate,
    /// High stress (`0.50..0.75`).
    High,
    /// Extreme stress (`>= 0.75`).
    Extreme,
}

/// SCR rate (events per minute).
#[must_use]
pub fn scr_rate(n_scr: usize, duration_s: f64) -> f64 {
    if duration_s <= 0.0 {
        return 0.0;
    }
    n_scr as f64 / (duration_s / 60.0)
}

/// Composite stress index from multiple physiological markers.
///
/// Combines SCR rate, mean SCL, and SCL variability into a single index.
#[must_use]
pub fn compute_stress_index(scr_rate_val: f64, mean_scl: f64, scl_variance: f64) -> f64 {
    0.4 * scr_rate_val.min(20.0) / 20.0
        + 0.3 * mean_scl.min(30.0) / 30.0
        + 0.3 * scl_variance.sqrt().min(5.0) / 5.0
}

/// Assess stress from EDA-derived features.
#[must_use]
pub fn assess_stress(scr_rate_val: f64, mean_scl: f64, scl_variance: f64) -> StressAssessment {
    let si = compute_stress_index(scr_rate_val, mean_scl, scl_variance);
    let level = match () {
        () if si < 0.25 => StressLevel::Low,
        () if si < 0.50 => StressLevel::Moderate,
        () if si < 0.75 => StressLevel::High,
        () => StressLevel::Extreme,
    };
    StressAssessment {
        stress_index: si,
        scr_rate: scr_rate_val,
        level,
    }
}

// ── Beat Classification ──────────────────────────────────────────────────────

/// Beat morphology class.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum BeatClass {
    /// Normal sinus rhythm.
    Normal,
    /// Premature ventricular contraction.
    PVC,
    /// Premature atrial contraction.
    PAC,
    /// Unclassified beat.
    Unknown,
}

/// Beat template for template-matching classification.
#[derive(Debug, Clone)]
pub struct BeatTemplate {
    /// Expected morphology class for this template.
    pub class: BeatClass,
    /// Template waveform (same length as beat windows).
    pub waveform: Vec<f64>,
}

/// Beat classification result.
#[derive(Debug, Clone)]
pub struct BeatResult {
    /// Best-matching template class.
    pub class: BeatClass,
    /// Normalized cross-correlation with the best template.
    pub correlation: f64,
}

/// Normalized cross-correlation between two equal-length signals.
///
/// # Panics
///
/// Panics if `a` and `b` have different lengths.
#[must_use]
pub fn normalized_correlation(a: &[f64], b: &[f64]) -> f64 {
    assert_eq!(a.len(), b.len());
    let n = a.len() as f64;
    let mean_a: f64 = a.iter().sum::<f64>() / n;
    let mean_b: f64 = b.iter().sum::<f64>() / n;

    let mut cov = 0.0;
    let mut var_a = 0.0;
    let mut var_b = 0.0;
    for (ai, bi) in a.iter().zip(b.iter()) {
        let da = ai - mean_a;
        let db = bi - mean_b;
        cov = da.mul_add(db, cov);
        var_a = da.mul_add(da, var_a);
        var_b = db.mul_add(db, var_b);
    }

    let denom = (var_a * var_b).sqrt();
    if denom < 1e-14 { 0.0 } else { cov / denom }
}

/// Classify a single beat window against a set of templates.
///
/// # Panics
/// Panics if `beat` and template waveforms have different lengths (skipped via `continue`).
#[must_use]
pub fn classify_beat(beat: &[f64], templates: &[BeatTemplate]) -> BeatResult {
    let mut best = BeatResult {
        class: BeatClass::Unknown,
        correlation: -1.0,
    };
    for tmpl in templates {
        if tmpl.waveform.len() != beat.len() {
            continue;
        }
        let corr = normalized_correlation(beat, &tmpl.waveform);
        if corr > best.correlation {
            best.correlation = corr;
            best.class = tmpl.class;
        }
    }
    best
}

/// Classify all beats in a batch.
#[must_use]
pub fn classify_all_beats(beats: &[Vec<f64>], templates: &[BeatTemplate]) -> Vec<BeatResult> {
    beats.iter().map(|b| classify_beat(b, templates)).collect()
}

/// Generate a synthetic normal sinus beat template.
#[must_use]
pub fn generate_normal_template(window_size: usize) -> Vec<f64> {
    let mut t = vec![0.0; window_size];
    let mid = window_size / 2;
    let sigma = window_size as f64 / 8.0;
    for (i, val) in t.iter_mut().enumerate() {
        let x = (i as f64 - mid as f64) / sigma;
        *val = (-0.5 * x * x).exp();
    }
    t
}

/// Generate a synthetic PVC (premature ventricular contraction) template.
#[must_use]
pub fn generate_pvc_template(window_size: usize) -> Vec<f64> {
    let mut t = vec![0.0; window_size];
    let mid = window_size / 2;
    let sigma = window_size as f64 / 6.0;
    for (i, val) in t.iter_mut().enumerate() {
        let x = (i as f64 - mid as f64) / sigma;
        *val = -1.2 * (-0.5 * x * x).exp();
    }
    t
}

/// Generate a synthetic PAC (premature atrial contraction) template.
#[must_use]
pub fn generate_pac_template(window_size: usize) -> Vec<f64> {
    let mut t = vec![0.0; window_size];
    let mid = window_size / 2;
    let sigma = window_size as f64 / 7.0;
    for (i, val) in t.iter_mut().enumerate() {
        let x = (i as f64 - mid as f64) / sigma;
        *val = 0.7 * (-0.5 * x * x).exp();
    }
    t
}

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

    #[test]
    fn test_eda_scl() {
        let signal = vec![1.0, 2.0, 3.0, 4.0, 5.0];
        let scl = eda_scl(&signal, 3);
        assert_eq!(scl.len(), 5);
        assert!((scl[0] - 2.0).abs() < 1e-10);
    }

    #[test]
    fn test_eda_phasic() {
        let signal = vec![1.0, 5.0, 1.0, 1.0, 1.0];
        let phasic = eda_phasic(&signal, 3);
        assert_eq!(phasic.len(), 5);
        assert!(phasic[1] > 0.0, "spike should produce phasic component");
    }

    #[test]
    fn test_scr_detection() {
        let phasic = vec![0.0, 0.1, 0.5, 0.3, 0.0, 0.0, 0.8, 0.2, 0.0];
        let peaks = eda_detect_scr(&phasic, 0.2);
        assert!(peaks.contains(&2));
        assert!(peaks.contains(&6));
    }

    #[test]
    fn test_stress_assessment() {
        let low = assess_stress(2.0, 5.0, 1.0);
        let high = assess_stress(15.0, 25.0, 20.0);
        assert!(low.stress_index < high.stress_index);
        assert_eq!(low.level, StressLevel::Low);
    }

    #[test]
    fn test_normalized_correlation_self() {
        let a = vec![1.0, 2.0, 3.0, 4.0, 5.0];
        let corr = normalized_correlation(&a, &a);
        assert!((corr - 1.0).abs() < 1e-10);
    }

    #[test]
    fn test_beat_classification() {
        let ws = 50;
        let normal = generate_normal_template(ws);
        let pvc = generate_pvc_template(ws);
        let templates = vec![
            BeatTemplate {
                class: BeatClass::Normal,
                waveform: normal.clone(),
            },
            BeatTemplate {
                class: BeatClass::PVC,
                waveform: pvc.clone(),
            },
        ];

        let result = classify_beat(&normal, &templates);
        assert_eq!(result.class, BeatClass::Normal);
        assert!(result.correlation > 0.9);

        let result_pvc = classify_beat(&pvc, &templates);
        assert_eq!(result_pvc.class, BeatClass::PVC);
    }

    #[test]
    fn test_generate_templates() {
        let n = generate_normal_template(50);
        let pvc = generate_pvc_template(50);
        let pac = generate_pac_template(50);
        assert_eq!(n.len(), 50);
        assert!(n[25] > 0.0);
        assert!(pvc[25] < 0.0);
        assert!(pac[25] > 0.0 && pac[25] < n[25]);
    }

    #[test]
    fn test_classify_all_beats() {
        let ws = 30;
        let normal = generate_normal_template(ws);
        let pvc = generate_pvc_template(ws);
        let templates = vec![
            BeatTemplate {
                class: BeatClass::Normal,
                waveform: normal.clone(),
            },
            BeatTemplate {
                class: BeatClass::PVC,
                waveform: pvc.clone(),
            },
        ];
        let beats = vec![normal, pvc];
        let results = classify_all_beats(&beats, &templates);
        assert_eq!(results[0].class, BeatClass::Normal);
        assert_eq!(results[1].class, BeatClass::PVC);
    }

    #[test]
    fn test_rolling_average_constant() {
        let signal = vec![5.0; 10];
        let avg = rolling_average(&signal, 3);
        assert_eq!(avg.len(), 10);
        for &v in &avg {
            assert!((v - 5.0).abs() < 1e-10);
        }
    }

    #[test]
    fn test_rolling_average_ramp() {
        let signal: Vec<f64> = (0..10).map(|i| i as f64).collect();
        let avg = rolling_average(&signal, 3);
        assert_eq!(avg.len(), 10);
        assert!((avg[1] - 1.0).abs() < 1e-10);
        assert!((avg[5] - 5.0).abs() < 1e-10);
    }

    #[test]
    fn test_convolve_1d_identity() {
        let signal = vec![1.0, 2.0, 3.0, 4.0, 5.0];
        let kernel = vec![1.0];
        let result = convolve_1d(&signal, &kernel);
        assert_eq!(result.len(), 5);
        assert!((result[2] - 3.0).abs() < 1e-10);
    }

    #[test]
    fn test_convolve_1d_averaging() {
        let signal = vec![0.0, 0.0, 1.0, 0.0, 0.0];
        let kernel = vec![1.0 / 3.0, 1.0 / 3.0, 1.0 / 3.0];
        let result = convolve_1d(&signal, &kernel);
        assert_eq!(result.len(), 3);
        assert!((result[1] - 1.0 / 3.0).abs() < 1e-10);
    }
}