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
// SPDX-License-Identifier: AGPL-3.0-or-later
//! Harmonic Angle Force (f64) — Three-Body Bonded Interaction
//!
//! **Potential**: U(θ) = ½`k_θ`(θ - θ₀)²
//! **Force**: Cartesian gradient of U over atoms i, j (vertex), k
//!
//! Three-body term used by all major biomolecular force fields.
//! Two-pass GPU dispatch: per-angle forces → reduce to per-particle.

use crate::device::WgpuDevice;
use crate::device::capabilities::{DeviceCapabilities, Fp64Strategy, WORKGROUP_SIZE_1D};
use crate::device::compute_pipeline::ComputeDispatch;
use crate::error::Result;
use std::sync::Arc;

/// f64 harmonic angle force calculator (GPU-accelerated).
pub struct HarmonicAngleF64 {
    device: Arc<WgpuDevice>,
}

/// Parameters for a single angle term (i-j-k, angle at j).
#[derive(Clone, Copy, Debug)]
pub struct HarmonicAngle {
    /// Particle index i
    pub i: u32,
    /// Vertex atom j (angle measured here)
    pub j: u32,
    /// Particle index k
    pub k: u32,
    /// Angular force constant `k_θ` (kJ/mol/rad²)
    pub force_constant: f64,
    /// Equilibrium angle θ₀ (radians)
    pub eq_angle: f64,
}

struct AngleBuffers {
    pos: wgpu::Buffer,
    triples: wgpu::Buffer,
    force_constants: wgpu::Buffer,
    eq_angles: wgpu::Buffer,
    angle_forces: wgpu::Buffer,
    params: wgpu::Buffer,
}

#[repr(C)]
#[derive(Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
struct Params {
    n_angles: u32,
    _p0: u32,
    _p1: u32,
    _p2: u32,
}

impl AngleBuffers {
    fn new(dev: &WgpuDevice, positions: &[f64], angles: &[HarmonicAngle]) -> Self {
        let n_angles = angles.len();

        let pos = dev
            .device
            .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                label: Some("angle pos"),
                contents: bytemuck::cast_slice(positions),
                usage: wgpu::BufferUsages::STORAGE,
            });

        let mut triple_data = Vec::with_capacity(n_angles * 3);
        let mut k_data = Vec::with_capacity(n_angles);
        let mut theta0_data = Vec::with_capacity(n_angles);
        for a in angles {
            triple_data.push(a.i);
            triple_data.push(a.j);
            triple_data.push(a.k);
            k_data.push(a.force_constant);
            theta0_data.push(a.eq_angle);
        }

        let triples = dev
            .device
            .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                label: Some("angle triples"),
                contents: bytemuck::cast_slice(&triple_data),
                usage: wgpu::BufferUsages::STORAGE,
            });

        let to_f64_buf = |label: &str, data: &[f64]| -> wgpu::Buffer {
            dev.device
                .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                    label: Some(label),
                    contents: bytemuck::cast_slice(data),
                    usage: wgpu::BufferUsages::STORAGE,
                })
        };

        let force_constants = to_f64_buf("angle k", &k_data);
        let eq_angles = to_f64_buf("angle theta0", &theta0_data);

        // 9 force components per angle (3 atoms × 3 dims)
        let angle_forces = dev.device.create_buffer(&wgpu::BufferDescriptor {
            label: Some("angle af"),
            size: (n_angles * 9 * 8) as u64,
            usage: wgpu::BufferUsages::STORAGE | wgpu::BufferUsages::COPY_SRC,
            mapped_at_creation: false,
        });

        let params = dev
            .device
            .create_buffer_init(&wgpu::util::BufferInitDescriptor {
                label: None,
                contents: bytemuck::bytes_of(&Params {
                    n_angles: n_angles as u32,
                    _p0: 0,
                    _p1: 0,
                    _p2: 0,
                }),
                usage: wgpu::BufferUsages::UNIFORM,
            });

        Self {
            pos,
            triples,
            force_constants,
            eq_angles,
            angle_forces,
            params,
        }
    }
}

fn reduce_angle_forces(
    dev: &WgpuDevice,
    shader_source: &str,
    angle_forces_buf: &wgpu::Buffer,
    triples_buf: &wgpu::Buffer,
    n_particles: usize,
    n_angles: usize,
) -> Result<Vec<f64>> {
    let particle_forces_buf = dev.device.create_buffer(&wgpu::BufferDescriptor {
        label: Some("angle pf"),
        size: (n_particles * 3 * 8) as u64,
        usage: wgpu::BufferUsages::STORAGE
            | wgpu::BufferUsages::COPY_SRC
            | wgpu::BufferUsages::COPY_DST,
        mapped_at_creation: false,
    });
    let mut enc = dev.create_encoder_guarded(&wgpu::CommandEncoderDescriptor { label: None });
    enc.clear_buffer(&particle_forces_buf, 0, None);
    dev.submit_commands(Some(enc.finish()));

    #[repr(C)]
    #[derive(Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
    struct ReduceParams {
        n_particles: u32,
        n_angles: u32,
        _p0: u32,
        _p1: u32,
    }
    let rp_buf = dev
        .device
        .create_buffer_init(&wgpu::util::BufferInitDescriptor {
            label: None,
            contents: bytemuck::bytes_of(&ReduceParams {
                n_particles: n_particles as u32,
                n_angles: n_angles as u32,
                _p0: 0,
                _p1: 0,
            }),
            usage: wgpu::BufferUsages::UNIFORM,
        });

    let wg = (n_particles as u32).div_ceil(WORKGROUP_SIZE_1D);
    ComputeDispatch::new(dev, "reduce_angle_forces_f64")
        .shader(shader_source, "reduce_angle_forces_f64")
        .f64()
        .uniform(0, &rp_buf)
        .storage_read(1, angle_forces_buf)
        .storage_read(2, triples_buf)
        .storage_rw(3, &particle_forces_buf)
        .dispatch(wg, 1, 1)
        .submit()?;

    dev.read_f64_buffer(&particle_forces_buf, n_particles * 3)
}

impl HarmonicAngleF64 {
    /// Create harmonic angle force calculator.
    /// # Errors
    /// Returns [`Err`] if device initialization fails.
    pub fn new(device: Arc<WgpuDevice>) -> Result<Self> {
        Ok(Self { device })
    }

    fn wgsl_shader() -> &'static str {
        include_str!("harmonic_angle_f64.wgsl")
    }

    fn wgsl_shader_for_device(device: &WgpuDevice) -> String {
        let caps = DeviceCapabilities::from_device(device);
        match caps.fp64_strategy() {
            Fp64Strategy::Sovereign | Fp64Strategy::Native | Fp64Strategy::Concurrent => {
                Self::wgsl_shader().to_string()
            }
            Fp64Strategy::Hybrid => Self::wgsl_shader().to_string(),
        }
    }

    /// Compute harmonic angle forces for all angles (GPU dispatch).
    /// # Errors
    /// Returns [`Err`] on GPU dispatch or readback failure.
    pub fn compute_forces(&self, positions: &[f64], angles: &[HarmonicAngle]) -> Result<Vec<f64>> {
        let n_particles = positions.len() / 3;
        if angles.is_empty() {
            return Ok(vec![0.0f64; n_particles * 3]);
        }
        self.compute_gpu(positions, angles, n_particles)
    }

    /// Compute harmonic angle forces and per-angle energies (GPU dispatch).
    /// # Errors
    /// Returns [`Err`] on GPU dispatch or readback failure.
    pub fn compute_forces_and_energy(
        &self,
        positions: &[f64],
        angles: &[HarmonicAngle],
    ) -> Result<(Vec<f64>, Vec<f64>)> {
        let n_particles = positions.len() / 3;
        if angles.is_empty() {
            return Ok((vec![0.0f64; n_particles * 3], vec![]));
        }
        self.compute_gpu_with_energy(positions, angles, n_particles)
    }

    fn compute_gpu(
        &self,
        positions: &[f64],
        angles: &[HarmonicAngle],
        n_particles: usize,
    ) -> Result<Vec<f64>> {
        let n_angles = angles.len();
        let dev = &self.device;
        let buffers = AngleBuffers::new(dev, positions, angles);

        let src = Self::wgsl_shader_for_device(dev);
        let wg = (n_angles as u32).div_ceil(WORKGROUP_SIZE_1D);
        ComputeDispatch::new(dev, "harmonic_angle_f64")
            .shader(&src, "harmonic_angle_f64")
            .f64()
            .storage_read(0, &buffers.pos)
            .storage_read(1, &buffers.triples)
            .storage_read(2, &buffers.force_constants)
            .storage_read(3, &buffers.eq_angles)
            .storage_rw(4, &buffers.angle_forces)
            .uniform(5, &buffers.params)
            .dispatch(wg, 1, 1)
            .submit()?;

        reduce_angle_forces(
            dev,
            &src,
            &buffers.angle_forces,
            &buffers.triples,
            n_particles,
            n_angles,
        )
    }

    fn compute_gpu_with_energy(
        &self,
        positions: &[f64],
        angles: &[HarmonicAngle],
        n_particles: usize,
    ) -> Result<(Vec<f64>, Vec<f64>)> {
        let n_angles = angles.len();
        let dev = &self.device;
        let buffers = AngleBuffers::new(dev, positions, angles);

        let energy_buf = dev.device.create_buffer(&wgpu::BufferDescriptor {
            label: Some("angle be"),
            size: (n_angles * 8) as u64,
            usage: wgpu::BufferUsages::STORAGE | wgpu::BufferUsages::COPY_SRC,
            mapped_at_creation: false,
        });

        let src = Self::wgsl_shader_for_device(dev);
        let wg = (n_angles as u32).div_ceil(WORKGROUP_SIZE_1D);
        ComputeDispatch::new(dev, "harmonic_angle_with_energy_f64")
            .shader(&src, "harmonic_angle_with_energy_f64")
            .f64()
            .storage_read(0, &buffers.pos)
            .storage_read(1, &buffers.triples)
            .storage_read(2, &buffers.force_constants)
            .storage_read(3, &buffers.eq_angles)
            .storage_rw(4, &buffers.angle_forces)
            .uniform(5, &buffers.params)
            .storage_rw(6, &energy_buf)
            .dispatch(wg, 1, 1)
            .submit()?;

        let energies = dev.read_f64_buffer(&energy_buf, n_angles)?;
        let forces = reduce_angle_forces(
            dev,
            &src,
            &buffers.angle_forces,
            &buffers.triples,
            n_particles,
            n_angles,
        )?;
        Ok((forces, energies))
    }
}

#[cfg(test)]
#[expect(
    clippy::suboptimal_flops,
    reason = "reference physics math in textbook notation for GPU kernel verification"
)]
mod tests {
    use super::*;

    fn compute_cpu(positions: &[f64], angles: &[HarmonicAngle]) -> Vec<f64> {
        let n_particles = positions.len() / 3;
        let mut forces = vec![0.0f64; n_particles * 3];
        for a in angles {
            let (i, j, k) = (a.i as usize, a.j as usize, a.k as usize);
            let rji = [
                positions[i * 3] - positions[j * 3],
                positions[i * 3 + 1] - positions[j * 3 + 1],
                positions[i * 3 + 2] - positions[j * 3 + 2],
            ];
            let rjk = [
                positions[k * 3] - positions[j * 3],
                positions[k * 3 + 1] - positions[j * 3 + 1],
                positions[k * 3 + 2] - positions[j * 3 + 2],
            ];
            let rji_len = (rji[0] * rji[0] + rji[1] * rji[1] + rji[2] * rji[2]).sqrt();
            let rjk_len = (rjk[0] * rjk[0] + rjk[1] * rjk[1] + rjk[2] * rjk[2]).sqrt();
            if rji_len < 1e-10 || rjk_len < 1e-10 {
                continue;
            }
            let cos_theta =
                (rji[0] * rjk[0] + rji[1] * rjk[1] + rji[2] * rjk[2]) / (rji_len * rjk_len);
            let cos_clamped = cos_theta.clamp(-1.0, 1.0);
            let theta = cos_clamped.acos();
            let sin_theta = (1.0 - cos_clamped * cos_clamped).sqrt().max(1e-12);

            let prefactor = -a.force_constant * (theta - a.eq_angle) / sin_theta;

            let rji_inv = 1.0 / rji_len;
            let rjk_inv = 1.0 / rjk_len;
            let rji_inv2 = rji_inv * rji_inv;
            let rjk_inv2 = rjk_inv * rjk_inv;

            for d in 0..3 {
                let fi = prefactor * (rjk[d] * rji_inv * rjk_inv - cos_clamped * rji[d] * rji_inv2);
                let fk = prefactor * (rji[d] * rji_inv * rjk_inv - cos_clamped * rjk[d] * rjk_inv2);
                forces[i * 3 + d] += fi;
                forces[k * 3 + d] += fk;
                forces[j * 3 + d] += -(fi + fk);
            }
        }
        forces
    }

    #[test]
    fn test_harmonic_angle_equilibrium_cpu() {
        let theta0: f64 = std::f64::consts::FRAC_PI_3 * 2.0; // 120°
        let r = 0.15_f64;
        let positions = vec![
            r * (theta0 / 2.0).cos(),
            r * (theta0 / 2.0).sin(),
            0.0,
            0.0,
            0.0,
            0.0,
            r * (theta0 / 2.0).cos(),
            -r * (theta0 / 2.0).sin(),
            0.0,
        ];
        let angles = vec![HarmonicAngle {
            i: 0,
            j: 1,
            k: 2,
            force_constant: 500.0,
            eq_angle: theta0,
        }];
        let forces = compute_cpu(&positions, &angles);
        for f in &forces {
            assert!(
                f.abs() < 1e-8,
                "force at equilibrium should be ~zero, got {f}"
            );
        }
    }

    #[test]
    fn test_harmonic_angle_momentum_conservation_cpu() {
        let positions = vec![0.15, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.15, 0.05];
        let angles = vec![HarmonicAngle {
            i: 0,
            j: 1,
            k: 2,
            force_constant: 500.0,
            eq_angle: std::f64::consts::FRAC_PI_2,
        }];
        let forces = compute_cpu(&positions, &angles);
        for d in 0..3 {
            let total = forces[d] + forces[3 + d] + forces[6 + d];
            assert!(
                total.abs() < 1e-10,
                "momentum not conserved on dim {d}: {total}"
            );
        }
    }
}