Hardware Reference
In-Depth Information
Table 5.3
Software version
Software
Version number
Linux
2.6.16
glibc
2.3.3
Table 5.4
Process-related performance
Performance
ratio (%)
TAS (us)
LL/SC (us)
Null call
0.44
0.44
100
Null I/O
0.83
0.83
100
Stat
12.20
8.69
140
Open/close
16.4
11.6
141
Select TCP
59.3
42.4
140
Sig inst
1.91
1.81
106
Sig handle
92.1
84.8
109
Fork proc
3,932
3,578
110
Exec proc
10,801
9,981
108
Shell proc
32 k
30 k
107
Table 5.5
Inter-process communication performance
Performance
ratio (%)
TAS (us)
LL/SC (us)
Pipe
17.5
14.3
122
AF UNIX
42
36.3
116
UDP
62.8
60.6
104
TCP
115
88.4
130
TCP conn
289
241
120
Table 5.6 File system and virtual memory related performance
Performance
ratio (%)
TAS (us)
LL/SC (us)
0 K fi le create
57.6
41.8
138
0 K fi le delete
32.6
24.4
134
10 K fi le create
281.3
238.5
118
10 K fi le delete
75.1
52.7
143
Mmap Latency
4,904
4,418
111
Page Fault
50
45.3
110
100fds select
39.5
25.1
157
The results are presented in Table 5.4 - 5.6 .
We can summarize that the SMP Linux using LL/SC atomic operation achieves
better performance than the SMP Linux using TAS atomic operation in all of the
LMBench results.
 
Search WWH ::




Custom Search