rustpower 0.5.0-rc.3

An experimental ECS world snapshot system built on Bevy, featuring structured archetype storage and manifest-based serialization.
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

#!/usr/bin/env python3
"""
scripts/plot_solve_breakdown.py

Per-solve cost breakdown: Assembly vs KLU-init vs KLU-iterative.

Three panels (IEEE 39, IEEE 118, PEGASE 9241).  Each panel shows V0/V1/V2
as horizontal stacked bars:

    Assembly   (dark, hatch)  — N_iter × per-iter assembly + symbolic build (V2)
    KLU init   (medium, none) — klu_l_analyze + klu_l_factor  (once per solve)
    KLU steady (light, none)  — (N_iter-1) × refactor + N_iter × back-sub

KLU init and KLU steady are version-invariant (same bars for V0/V1/V2);
only the Assembly bar shrinks V0→V2.

Data source: paper/solve_breakdown.csv (written by bench_jacobian_fill Rust test).

Usage:
    python scripts/plot_solve_breakdown.py
"""

import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import pandas as pd
from pathlib import Path

df = pd.read_csv('paper/solve_breakdown.csv')

SYSTEMS   = ['IEEE 39', 'IEEE 118', 'PEGASE 9241']
VERSIONS  = ['V0', 'V1', 'V2']
SYS_TITLE = {'IEEE 39': 'IEEE 39', 'IEEE 118': 'IEEE 118', 'PEGASE 9241': 'PEGASE 9241'}

VER_LABEL_FULL  = {
    'V0': 'V0  (MATPOWER)',
    'V1': "V1  (Sch\xe4fer '18)",
    'V2': 'V2  (this work)',
}
VER_LABEL_SHORT = {'V0': 'V0', 'V1': 'V1', 'V2': 'V2'}

# ─── Styles ────────────────────────────────────────────────────────────────────
STYLE_ASM  = dict(facecolor='0.20', hatch='////', edgecolor='white',  linewidth=0.0)
STYLE_INIT = dict(facecolor='0.55', hatch='',     edgecolor='0.40',   linewidth=0.6)
STYLE_ITER = dict(facecolor='0.82', hatch='',     edgecolor='0.40',   linewidth=0.6)

plt.rcParams.update({
    'font.family':       'serif',
    'font.size':         8,
    'axes.linewidth':    0.55,
    'xtick.major.width': 0.55,
    'ytick.major.width': 0.55,
    'xtick.major.size':  3,
    'hatch.linewidth':   0.55,
})

fig, axes = plt.subplots(1, 3, figsize=(7.1, 2.6))
fig.subplots_adjust(wspace=0.32, bottom=0.35)

BAR_H = 0.50
Y = {'V0': 2, 'V1': 1, 'V2': 0}

for idx, (ax, sys_name) in enumerate(zip(axes, SYSTEMS)):
    sys_df = df[df['system'] == sys_name].set_index('version')

    # Per-solve segment totals (ms) for each version.
    segs = {}
    for ver in VERSIONS:
        row = sys_df.loc[ver]
        n   = row['n_iter']
        # Assembly: N×per-iter + one-time symbolic build (sym_us=0 for V0/V1).
        asm_ms   = (n * row['asm_us'] + row['sym_us']) / 1e3
        # KLU one-time per solve: analyze + first full factor.
        init_ms  = (row['analyze_us'] + row['factor_us']) / 1e3
        # KLU iterative: (N-1) refactors + N back-subs.
        iter_ms  = ((n - 1) * row['refactor_us'] + n * row['backsolve_us']) / 1e3
        segs[ver] = (asm_ms, init_ms, iter_ms)

    x_max = max(sum(s) for s in segs.values())

    for ver in VERSIONS:
        y              = Y[ver]
        asm, init, itr = segs[ver]
        total          = asm + init + itr

        # Assembly bar.
        ax.barh(y, asm, height=BAR_H, left=0, zorder=3, **STYLE_ASM)
        # KLU init bar.
        ax.barh(y, init, height=BAR_H, left=asm, zorder=3, **STYLE_INIT)
        # KLU steady bar.
        ax.barh(y, itr, height=BAR_H, left=asm + init, zorder=3, **STYLE_ITER)
        # Outer border.
        ax.barh(y, total, height=BAR_H,
                facecolor='none', edgecolor='0.30', linewidth=0.6, zorder=4)

        # Assembly % label (white, inside bar, only when wide enough).
        pct = asm / total * 100
        if asm > x_max * 0.08:
            ax.text(asm / 2, y, f'{pct:.0f}%',
                    ha='center', va='center', fontsize=5.8,
                    color='white', fontweight='bold', zorder=5)

    ax.set_title(SYS_TITLE[sys_name], fontsize=8.5, fontweight='bold', pad=3)
    ax.set_yticks([0, 1, 2])
    if idx == 0:
        ax.set_yticklabels(
            [VER_LABEL_FULL['V2'], VER_LABEL_FULL['V1'], VER_LABEL_FULL['V0']],
            fontsize=6.8)
    else:
        ax.set_yticklabels(
            [VER_LABEL_SHORT['V2'], VER_LABEL_SHORT['V1'], VER_LABEL_SHORT['V0']],
            fontsize=7)
    ax.set_xlabel('Per-solve time (ms)', fontsize=8)
    ax.tick_params(axis='x', labelsize=7)
    ax.set_ylim(-0.55, 2.55)
    ax.set_xlim(0, x_max * 1.08)
    ax.grid(axis='x', linestyle=':', linewidth=0.4, color='0.65', zorder=0)
    ax.spines['top'].set_visible(False)
    ax.spines['right'].set_visible(False)

# ─── Legend ────────────────────────────────────────────────────────────────────
h_asm  = mpatches.Patch(facecolor='0.20', hatch='////', edgecolor='0.40',
                         linewidth=0.55, label='Jacobian assembly  (N \xd7 per-iter + symbolic build)')
h_init = mpatches.Patch(facecolor='0.55', hatch='',    edgecolor='0.40',
                         linewidth=0.55, label='KLU init: analyze + factor  (once per solve, version-invariant)')
h_iter = mpatches.Patch(facecolor='0.82', hatch='',    edgecolor='0.40',
                         linewidth=0.55, label='KLU iterative: re-factor + back-sub  (version-invariant)')

fig.legend(handles=[h_asm, h_init, h_iter],
           loc='lower center', ncol=1, fontsize=6.5,
           frameon=False, bbox_to_anchor=(0.5, 0.01),
           handlelength=2.4, handletextpad=0.6)

fig.text(
    0.015, -0.02,
    'Per-solve time reconstructed from directly instrumented component averages '
    '(bench_jacobian_fill, export_solve_breakdown_csv).  '
    'KLU segments are identical for V0/V1/V2.',
    fontsize=5.5, color='0.45')

# ─── Save ──────────────────────────────────────────────────────────────────────
out = Path('paper/solve_breakdown')
fig.savefig(str(out) + '.pdf', bbox_inches='tight', dpi=300)
fig.savefig(str(out) + '.png', bbox_inches='tight', dpi=300)
print(f'Saved {out}.pdf and {out}.png')

# ─── Console summary ───────────────────────────────────────────────────────────
print()
print('Assembly share of estimated per-solve time:')
for sys_name in SYSTEMS:
    sys_df = df[df['system'] == sys_name].set_index('version')
    parts = {}
    for ver in VERSIONS:
        row = sys_df.loc[ver]
        n   = row['n_iter']
        asm = (n * row['asm_us'] + row['sym_us']) / 1e3
        tot = asm + (row['analyze_us'] + row['factor_us']) / 1e3 \
                  + ((n-1)*row['refactor_us'] + n*row['backsolve_us']) / 1e3
        parts[ver] = asm / tot * 100
    print(f'  {sys_name}: V0={parts["V0"]:.0f}%  V1={parts["V1"]:.0f}%  V2={parts["V2"]:.0f}%')