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
|
import matplotlib.pyplot as plt, csv, sys, numpy as np
if len(sys.argv) < 2:
print("Please provide path to input data file.")
sys.exit()
read = {}
labels = {}
cols = 0
rows = 0
with open(sys.argv[1]) as datafile:
reader = csv.reader(datafile)
headers = reader.__next__()
cols = len(headers)
for i in range(cols):
read[i] = []
labels[i] = headers[i]
for row in reader:
rows += 1
for i in range(cols):
read[i].append(row[i])
data = {}
for i in range(cols):
data[labels[i]] = read[i]
def plot_point_count():
for method in ["Inside", "DistanceToIn", "SafetyToIn"]:
fig = plt.figure()
ax = fig.add_subplot(111)
ax.set_xscale("log")
point_count = []
elapsed_scalar = []
elapsed_vector = []
elapsed_cuda = []
elapsed_cudamem = []
for i in range(rows):
if data["method"][i] == method:
if data["library"][i] == "Specialized":
point_count.append(int(data["points"][i]))
elapsed_scalar.append(float(data["elapsed"][i])/point_count[-1])
if data["library"][i] == "Vectorized":
elapsed_vector.append(elapsed_scalar[-1]/(float(data["elapsed"][i])
/point_count[-1]))
if data["library"][i] == "CUDA":
elapsed_cuda.append(elapsed_scalar[-1]/(float(data["elapsed"][i])
/point_count[-1]))
if data["library"][i] == "CUDAMemory":
elapsed_cudamem.append(elapsed_scalar[-1]/(float(data["elapsed"][i])
/point_count[-1]))
ax.plot(point_count, np.ones(len(point_count), dtype=np.float), "--b",
label="Scalar")
ax.plot(point_count, elapsed_vector, "-xr", ms=5,
label="CPU SIMD instructions")
ax.plot(point_count, elapsed_cuda, "-xg", ms=5,
label="CUDA, best case memory overhead")
ax.plot(point_count, elapsed_cudamem, "--", color=[0, 1, 0], ms=5,
label="CUDA, worst case memory overhead")
ax.set_xlim(point_count[0], point_count[-1])
plt.xticks(point_count, [str(x) for x in point_count])
ybegin, yend = ax.get_ylim()
ax.set_yticks(np.arange(0, yend, 1.0))
ax.set_title("Performance of %s algorithm for tubes" % method)
ax.set_xlabel("Input length")
ax.set_ylabel("Speedup")
ax.legend(loc="upper left")
fig.savefig("%s.pdf" % method)
plot_point_count()
|
