Question

I have 2 python scripts https://gist.github.com/2233477.

• rsgen.py generates “random” inputs for use in simulate.py
• simulate.py does the actual simulation

Thing is, when I start to increase the input size from rsgen.py with the --numReferences param, I get different outputs

# ./rsgen.py --numReferences 1000 > rs.txt; cat rs.txt | xargs ./simulate.py
Number of page faults : 59

# ./rsgen.py --numReferences 100000 > rs.txt; cat rs.txt | xargs ./simulate.py
Number of page faults : 873
Number of page faults : 848
Number of page faults : 823
Number of page faults : 103

./rsgen.py --numReferences 1000000 > rs.txt; cat rs.txt | xargs ./simulate.py
Number of page faults : 866
Number of page faults : 869
Number of page faults : 876
Number of page faults : 907
Number of page faults : 910
Number of page faults : 1001
Number of page faults : 845
...

Notice as I increase numReferences, the python script simulate appears to run more times. Why is that? I am expecting just 1 line of “Number of page faults: …”

Answer 1

This probably has something to do with xargs‘ ARG_MAX which defines a batch size for how many args to send to an executable; hence why multiple invocations of your script since it is splitting up the args across multiple calls.

Try the -n (or --max-args) flag of xargs .

A better way alltogether would be to have simulate.py accept a file argument so you could do something like this:

./rsgen.py --numReferences N > rs.txt; 
./simulate.py -f rs.txt

It would probably be a lot faster since it avoids the xargs overhead