forked from jmduarte/a3d3_graph
-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmake_sciml_xdr_graph.py
129 lines (119 loc) · 4.39 KB
/
make_sciml_xdr_graph.py
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
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import mplhep as hep
import numpy as np
import math
# label: data rate lower bound [B/s], data rate upper bound [B/s], latency lower bound [s], latency upper bound [s]
input_dict = {
"LHC sensor": [25e12, 100e12, 10e-9, 25e-9],
# "LHC near-sensor": [48 * 40e6, 48 * 40e6, 25e-9, 100e-9],
"LHC trigger": [32 * 40e6, 32 * 40e6, 100e-9, 5e-6],
# "Beam Control": [3e3 * 15, 3e3 * 15, 100e-6, 5e-3], # Booster
"Beam control": [3e3 * 15, 3e3 * 15, 100e-6, 5e-3], # Booster control
"Magnet quench": [3e6, 3e6, 100e-6, 100e-6], # Quench
# "Pixel": [1.28e9/8 * 40/.75, 4*1.28e9/8 * 40/.75, 10e-9, 25e-9],
# "DUNE": [1e9, 10e9, 1, 5*60],
"DUNE readout": [0.8e9, 0.8e9, 1e-6, 1e-6],
# "Quantum": [1e9, 7e9, 500e-9, 1000e-9],
# "Quantum": [40e9, 40e9, 100e-9, 1e-6],
"Qubit Readout": [9e9, 9e9, 100e-9, 100e-9],
"EIC trigger": [7.5e9, 7.5e9, 500e-9, 500e-9],
"X-ray diffraction": [10e6, 100e6, 1e-6, 20e-6],
"Electron microscopy": [0.6e9, 0.6e9, 50e-6, 50e-6],
"Plasma control": [3e9, 3e9, 5e-6, 20e-6],
"Neuro": [5e6, 5e6, 1e-3, 1e-3],
"Internet-of-things": [3e3 / 100e-3, 3e3 / 1e-3, 1e-3, 100e-3],
"Mobile devices": [1e3 / 100e-3, 1e3 / 40e-3, 40e-3, 100e-3],
}
labels = input_dict.keys()
ylo = np.array([input_dict[key][0] for key in labels])
yhi = np.array([input_dict[key][1] for key in labels])
xlo = np.array([input_dict[key][2] for key in labels])
xhi = np.array([input_dict[key][3] for key in labels])
colors = [
"#1f77b4",
"#ff7f0e",
"#2ca02c",
"#9467bd",
"#8c564b",
"#e377c2",
"#17becf",
"#7f7f7f",
"#bcbd22",
"#d62728",
]
### get more
colors += ["#1f77b4","#9467bd","#bcbd22","#17becf","#ff7f0e",]
colors += ["#bbbbbb"]*10
plt.style.use([hep.style.ROOT, hep.style.firamath])
ymin = 1e2
ymax = 2e14
xmin = 1e-9
xmax = 1e5
f, ax = plt.subplots()
# FastML contour
#ax.text(2e-9, 2e10, "FastML Science (WIP)", color="gray", style="italic", weight="bold")
ax.text(2e-2, 2e13, "Fast ML for Science", color="gray", style="italic", weight="bold")
ax.text(2e-2, 5e12, "benchmark tasks", color="gray", style="italic", fontsize=22)
box_y = np.array([3e3 * 15, 3e3 * 15, ymax, ymax])
box_x = np.array([xmin, 5e-3, 5e-3, xmin])
ax.fill(box_x, box_y, "gray", alpha=0.2)
for xloi, xhii, yloi, yhii, l, c in zip(xlo, xhi, ylo, yhi, labels, colors):
yi = math.sqrt(yloi * yhii)
xi = math.sqrt(xloi * xhii)
ax.errorbar(
[xi],
[yi],
yerr=[[yi - yloi], [yhii - yi]],
xerr=[[xi - xloi], [xhii - xi]],
label=l,
marker="",
capsize=6,
markersize=10,
color=c,
)
sz=20.
if "Internet-of-things" in l:
ax.text(xi * 5, yi * 2, l, color=c, size=sz)
elif "Mobile devices" in l:
ax.text(xi * 2, yi * 2, l, color=c, size=sz)
elif "Beam control" in l:
ax.text(xi / 5e3, yi * 2, l, color=c, size=sz)
elif "LHC sensor" in l:
ax.text(xi * 3, yi / 2, l, color=c, size=sz)
elif "LHC near-sensor" in l:
ax.text(xi / 20, yi * 3.5, "LHC", color=c, size=sz)
ax.text(xi / 20, yi * 1.5, "near-sensor", color=c, size=sz)
elif "X-ray diffraction" in l:
ax.text(xi * 3, yi * 1.6, "X-ray diffraction", color=c, size=sz)
# ax.text(xi / 2e3, yi / 2, "(BraggNN)", color=c, size=sz)
elif "EIC trigger" in l:
ax.text(xi / 100, yi / 3, l, color=c, size=sz)
elif "Qubit Readout" in l:
ax.text(xi / 10, yi * 1.9, l, color=c, size=sz)
elif "Plasma control" in l:
ax.text(xi * 3, yi * 1, l, color=c, size=sz)
elif "DUNE readout" in l:
ax.text(xi / 5e2, yi / 3, l, color=c, size=sz)
elif "DUNE" in l:
ax.text(xi / 1e2, yi / 6, l, color=c, size=sz)
elif "LHC trigger" in l:
ax.text(xi * 10, yi / 1.2, l, color=c, size=sz)
elif "Magnet quench" in l:
ax.text(xi / 1e3, yi / 4, l, color=c, size=sz)
elif "Electron microscopy" in l:
ax.text(xi * 1.5, yi / 2, l, color=c, size=sz)
elif "Neuro" in l:
ax.text(xi / 5, yi * 1.7, l, color=c, size=sz)
else:
ax.text(xi * 1, yi * 1, l, color=c, size=sz)
# ax.text(xi * 2, yi * 2, l, color=c, size=10.)
ax.loglog()
ax.set_xlim(xmin, xmax)
ax.set_ylim(ymin, ymax)
ax.set_xlabel("Computation time [s]")
ax.set_ylabel("Data rate [B/s]")
plt.tight_layout()
plt.savefig("sciml_graph.pdf")
plt.savefig("sciml_graph.png")