Mixed-Language Programming with XcalableMP
161
1
2
4 10 20
Number of XMP nodes
1000
100
10
1
0.1
Execution Time (Sec.)
40 80 160 320 640 1280
Parallel Efficiency
Execution Time
1.00
0.75
0.50
0.25
0.00
Parallel Efficiency
1.55
0.40
0.21
0.13
3.09
12.33
24.64
123.17
61.60
0.78
6.17
Fig. 11 Performance evaluation
100
90
80
70
60
50
40
30
20
10
0
40 80 160 320 640 1280
Number of nodes
Communication
Calculation
1
2
4 10 20
Calc./Comm. Ratio (%)
Fig. 12 Calculation and communication ratio
5 Conclusion
This chapter describes how to use linkage functions between XMP and MPI library
in Omni compiler. Moreover, it also describes how to call an XMP program from a
Python program. Users can call functions written in these languages with a simple