Command Processor Scheduler

Usage

Execute shell commands in parallel.

mpirun -np num_procs ./cps tasklist

For example,

make
mpirun -np 4 ./cps task.sh

The shell commands are given by list.

$ cat task.sh
# tasks
sleep 1.1
sleep 1.4
sleep 0.9
sleep 1.1
sleep 1.3
sleep 1.2

The lines beginning with # are considered to be comment.

It will generate a log file cps.log after execution.

$ cat cps.log
Number of tasks : 6
Number of processes : 4
Total execution time: 7.043 [s]
Elapsed time: 2.613 [s]
Parallel Efficiency : 0.898456

Task list:
Command : Elapsed time
sleep 1.1 : 1.11 [s]
sleep 1.4 : 1.408 [s]
sleep 0.9 : 0.907 [s]
sleep 1.1 : 1.106 [s]
sleep 1.3 : 1.308 [s]
sleep 1.2 : 1.204 [s]

One process is used for scheduler. Therefore, the number of processes to execute tasks is num_procs - 1. In this case, the total execution time is 7.043 s and the elapsed time (wall time) is 2.613 s. The effective number of processes is 3 (4-1).

Therefore, the parallel efficiency for this case is

7.043 / (2.613 * 3) = 0.8984564357698686 ~ 89.8%

License

This software is released under the MIT License, see LICENSE.