rustmatrix 2.1.1

Rust-backed T-matrix scattering for nonspherical particles (port of pytmatrix)
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

"""Speed comparison: rustmatrix (Rust) vs pytmatrix (Fortran).

Run in an env where both are importable:
    conda run -n rupy-parity python benches/bench_vs_pytmatrix.py
"""

from __future__ import annotations

import time

import numpy as np
import pytmatrix.tmatrix as py_tm
from pytmatrix import orientation as py_orient
from pytmatrix import psd as py_psd

import rustmatrix
from rustmatrix import Scatterer, orientation as rs_orient, psd as rs_psd
from rustmatrix.tmatrix_aux import geom_horiz_back


def _time(fn, repeats):
    # Discard one warmup call so cold caches don't contaminate the timing.
    fn()
    t = time.perf_counter()
    for _ in range(repeats):
        fn()
    return (time.perf_counter() - t) / repeats


def bench_single_orientation():
    """Cold T-matrix build + amplitude matrix, fresh Scatterer each call."""
    geom = geom_horiz_back
    kwargs = dict(radius=1.0, wavelength=6.283185307, axis_ratio=1.5,
                  m=complex(1.5, 0.01), ddelt=1e-4, ndgs=2)

    def py_fn():
        s = py_tm.Scatterer(**kwargs)
        s.set_geometry(geom)
        s.get_SZ()

    def rs_fn():
        s = Scatterer(**kwargs)
        s.set_geometry(geom)
        s.get_SZ()

    return _time(py_fn, 20), _time(rs_fn, 20)


def bench_cached_evaluate():
    """Warm cache — just re-evaluate S/Z at a new orientation."""
    geom = geom_horiz_back
    kwargs = dict(radius=1.0, wavelength=6.283185307, axis_ratio=1.5,
                  m=complex(1.5, 0.01), ddelt=1e-4, ndgs=2)

    py = py_tm.Scatterer(**kwargs)
    py.set_geometry(geom)
    py.get_SZ()  # prime

    rs = Scatterer(**kwargs)
    rs.set_geometry(geom)
    rs.get_SZ()

    def py_fn():
        # Vary alpha so scatter cache misses but T-matrix cache stays warm.
        py.alpha = (py.alpha + 7.0) % 360.0
        py.get_SZ()

    def rs_fn():
        rs.alpha = (rs.alpha + 7.0) % 360.0
        rs.get_SZ()

    return _time(py_fn, 200), _time(rs_fn, 200)


def bench_orient_averaged_fixed():
    """Gaussian-PDF orientation averaging, 4 alpha × 8 beta = 32 orientations."""
    kwargs = dict(radius=1.0, wavelength=6.283185307, axis_ratio=2.0,
                  m=complex(1.5, 0.01), ddelt=1e-4, ndgs=2)

    def py_fn():
        s = py_tm.Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.or_pdf = py_orient.gaussian_pdf(std=20.0, mean=90.0)
        s.orient = py_orient.orient_averaged_fixed
        s.n_alpha = 4
        s.n_beta = 8
        s.get_SZ()

    def rs_fn():
        s = Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.or_pdf = rs_orient.gaussian_pdf(std=20.0, mean=90.0)
        s.orient = rs_orient.orient_averaged_fixed
        s.n_alpha = 4
        s.n_beta = 8
        s.get_SZ()

    return _time(py_fn, 5), _time(rs_fn, 5)


def bench_psd_tabulate_orient_avg(num_points):
    """Orientation-averaged (n_alpha=4, n_beta=8) PSD tabulation."""
    kwargs = dict(wavelength=6.283185307, m=complex(1.5, 0.01), axis_ratio=2.0,
                  ddelt=1e-4, ndgs=2)

    def py_fn():
        s = py_tm.Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.or_pdf = py_orient.gaussian_pdf(std=20.0, mean=90.0)
        s.orient = py_orient.orient_averaged_fixed
        s.n_alpha = 4
        s.n_beta = 8
        s.psd_integrator = py_psd.PSDIntegrator()
        s.psd_integrator.num_points = num_points
        s.psd_integrator.D_max = 4.0
        s.psd = py_psd.GammaPSD(D0=1.0, Nw=1e3, mu=4)
        s.psd_integrator.init_scatter_table(s)

    def rs_fn():
        s = Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.or_pdf = rs_orient.gaussian_pdf(std=20.0, mean=90.0)
        s.orient = rs_orient.orient_averaged_fixed
        s.n_alpha = 4
        s.n_beta = 8
        s.psd_integrator = rs_psd.PSDIntegrator()
        s.psd_integrator.num_points = num_points
        s.psd_integrator.D_max = 4.0
        s.psd = rs_psd.GammaPSD(D0=1.0, Nw=1e3, mu=4)
        s.psd_integrator.init_scatter_table(s)

    return _time(py_fn, 2), _time(rs_fn, 2)


def bench_psd_tabulate(num_points):
    """Tabulate S,Z at `num_points` diameters (dominant cost of a PSD run)."""
    kwargs = dict(wavelength=6.5, m=complex(1.5, 0.5), axis_ratio=1.0,
                  ddelt=1e-4, ndgs=2)

    def py_fn():
        s = py_tm.Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.psd_integrator = py_psd.PSDIntegrator()
        s.psd_integrator.num_points = num_points
        s.psd_integrator.D_max = 10.0
        s.psd = py_psd.GammaPSD(D0=1.0, Nw=1e3, mu=4)
        s.psd_integrator.init_scatter_table(s)

    def rs_fn():
        s = Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.psd_integrator = rs_psd.PSDIntegrator()
        s.psd_integrator.num_points = num_points
        s.psd_integrator.D_max = 10.0
        s.psd = rs_psd.GammaPSD(D0=1.0, Nw=1e3, mu=4)
        s.psd_integrator.init_scatter_table(s)

    return _time(py_fn, 3), _time(rs_fn, 3)


def bench_psd_tabulate_angular(num_points):
    """PSD tabulation with angular_integration=True (sca_xsect/ext_xsect/asym)."""
    kwargs = dict(wavelength=6.5, m=complex(1.5, 0.5), axis_ratio=1.0,
                  ddelt=1e-4, ndgs=2)

    def py_fn():
        s = py_tm.Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.psd_integrator = py_psd.PSDIntegrator()
        s.psd_integrator.num_points = num_points
        s.psd_integrator.D_max = 10.0
        s.psd = py_psd.GammaPSD(D0=1.0, Nw=1e3, mu=4)
        s.psd_integrator.init_scatter_table(s, angular_integration=True)

    def rs_fn():
        s = Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.psd_integrator = rs_psd.PSDIntegrator()
        s.psd_integrator.num_points = num_points
        s.psd_integrator.D_max = 10.0
        s.psd = rs_psd.GammaPSD(D0=1.0, Nw=1e3, mu=4)
        s.psd_integrator.init_scatter_table(s, angular_integration=True)

    return _time(py_fn, 2), _time(rs_fn, 2)


def bench_psd_tabulate_orient_adaptive(num_points):
    """PSD tabulation with orient_averaged_adaptive (scipy.dblquad on pytmatrix side)."""
    kwargs = dict(wavelength=6.283185307, m=complex(1.5, 0.01), axis_ratio=1.5,
                  ddelt=1e-4, ndgs=2)

    def py_fn():
        s = py_tm.Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.or_pdf = py_orient.gaussian_pdf(std=20.0, mean=90.0)
        s.orient = py_orient.orient_averaged_adaptive
        s.psd_integrator = py_psd.PSDIntegrator()
        s.psd_integrator.num_points = num_points
        s.psd_integrator.D_max = 3.0
        s.psd = py_psd.GammaPSD(D0=1.0, Nw=1e3, mu=4)
        s.psd_integrator.init_scatter_table(s)

    def rs_fn():
        s = Scatterer(**kwargs)
        s.set_geometry(geom_horiz_back)
        s.or_pdf = rs_orient.gaussian_pdf(std=20.0, mean=90.0)
        s.orient = rs_orient.orient_averaged_adaptive
        s.psd_integrator = rs_psd.PSDIntegrator()
        s.psd_integrator.num_points = num_points
        s.psd_integrator.D_max = 3.0
        s.psd = rs_psd.GammaPSD(D0=1.0, Nw=1e3, mu=4)
        s.psd_integrator.init_scatter_table(s)

    return _time(py_fn, 1), _time(rs_fn, 2)


def bench_tmatrix_only():
    """Just the T-matrix computation (CALCTMAT), no amplitude evaluation."""
    radius, wl = 1.5, 6.283185307
    m = complex(1.5, 0.01)

    # pytmatrix wraps CALCTMAT inside Scatterer._init_tmatrix; easiest apples-
    # to-apples is to build a Scatterer and never evaluate SZ.
    def py_fn():
        s = py_tm.Scatterer(radius=radius, wavelength=wl, axis_ratio=1.5,
                            m=m, ddelt=1e-4, ndgs=2)
        s._init_tmatrix()

    def rs_fn():
        rustmatrix.calctmat(radius, 1.0, wl, m.real, m.imag, 1.5, -1, 1e-4, 2)

    return _time(py_fn, 20), _time(rs_fn, 20)


def main():
    cases = [
        ("calctmat only (spheroid ax=1.5)", bench_tmatrix_only),
        ("single orient, cold (fresh Scatterer)", bench_single_orientation),
        ("cached re-eval (warm T-matrix)", bench_cached_evaluate),
        ("orient-averaged fixed (4×8 = 32)", bench_orient_averaged_fixed),
        ("PSD init_scatter_table, 32 points", lambda: bench_psd_tabulate(32)),
        ("PSD init_scatter_table, 64 points", lambda: bench_psd_tabulate(64)),
        ("PSD + orient-avg (4×8), 32 points",
         lambda: bench_psd_tabulate_orient_avg(32)),
        ("PSD + orient-avg (4×8), 64 points",
         lambda: bench_psd_tabulate_orient_avg(64)),
        ("PSD + angular_integration, 32 points",
         lambda: bench_psd_tabulate_angular(32)),
        ("PSD + angular_integration, 64 points",
         lambda: bench_psd_tabulate_angular(64)),
        ("PSD + orient-adaptive, 4 points",
         lambda: bench_psd_tabulate_orient_adaptive(4)),
    ]

    print(f"{'case':<44} {'pytmatrix':>12} {'rustmatrix':>14} {'speedup':>10}")
    print("-" * 82)
    for name, fn in cases:
        t_py, t_rs = fn()
        ratio = t_py / t_rs if t_rs > 0 else float("inf")
        arrow = "faster" if ratio >= 1 else "slower"
        print(f"{name:<44} {t_py*1000:>10.2f}ms {t_rs*1000:>12.2f}ms "
              f"{ratio:>7.2f}× {arrow}")


if __name__ == "__main__":
    main()