Bug 1578

Summary:	Excessive execution time that varies with the setting m_step
Product:	ns-3	Reporter:	Julian <julian.quindt>
Component:	mesh	Assignee:	Kirill Andreev <andreev>
Status:	NEEDINFO ---
Severity:	normal	CC:	ns-bugs, tomh
Priority:	P5
Version:	ns-3.16
Hardware:	All
OS:	All

Description Julian 2013-02-04 08:30:14 EST

With 3.16 and 3.15, I observe a massive variation of the execution time ranging from reasonable to excessive (3.14.1 did not show that issue). Furthermore, the execution time seems to develop quite chaotic when changing the m_step parameter. The excessive runs are totally reproducible.

Here is what you can do to reproduce:
1.) Use 3.15 or 3.16, of the shelf
2.) Use the mesh.cc from src/mesh/examples
3.) Set m_xSize=4, m_ySize=4, m_totalTime=1000
4.) Vary the m_step parameter (do 50 meters first, thats fast, do 30 meters then, thats excessive)

Voilá.
Hope to get a comment on this one. Thank you. Sincerely. Julian

Here are execution times I saw:
m_step  Execution Time  
100	17,122
90	17,343
80	33,129
70	52,942
60	155,196
50	52,808
40	121,793
30	5009,038 <----------
20	74,852
10	88,265

Comment 1 Julian 2013-02-04 08:31:51 EST

Addendum: I saw this behavior on the latest Ubuntu i386 and on Mountain Lion 10.8.3.

Comment 2 Kirill Andreev 2013-04-14 08:48:06 EDT

(In reply to comment #0)
> With 3.16 and 3.15, I observe a massive variation of the execution time ranging
> from reasonable to excessive (3.14.1 did not show that issue). Furthermore, the
> execution time seems to develop quite chaotic when changing the m_step
> parameter. The excessive runs are totally reproducible.
> 
> Here is what you can do to reproduce:
> 1.) Use 3.15 or 3.16, of the shelf
> 2.) Use the mesh.cc from src/mesh/examples
> 3.) Set m_xSize=4, m_ySize=4, m_totalTime=1000
> 4.) Vary the m_step parameter (do 50 meters first, thats fast, do 30 meters
> then, thats excessive)
> 
> Voilá.
> Hope to get a comment on this one. Thank you. Sincerely. Julian
> 
> Here are execution times I saw:
> m_step  Execution Time  
> 100    17,122
> 90    17,343
> 80    33,129
> 70    52,942
> 60    155,196
> 50    52,808
> 40    121,793
> 30    5009,038 <----------
> 20    74,852
> 10    88,265

What RngRun did you use? Which build? I can not reproduce this in the current ns-3-dev
$ time ./waf --run "mesh --step=30 --x-size=4, y-size=4, --time=1000"
Waf: Entering directory `/Volumes/Documents/Work/ns-3-dev/build'
Waf: Leaving directory `/Volumes/Documents/Work/ns-3-dev/build'
'build' finished successfully (1.529s)
Printing mesh point device #0 diagnostics to mp-report-0.xml
Printing mesh point device #1 diagnostics to mp-report-1.xml
Printing mesh point device #2 diagnostics to mp-report-2.xml
Printing mesh point device #3 diagnostics to mp-report-3.xml
Printing mesh point device #4 diagnostics to mp-report-4.xml
Printing mesh point device #5 diagnostics to mp-report-5.xml
Printing mesh point device #6 diagnostics to mp-report-6.xml
Printing mesh point device #7 diagnostics to mp-report-7.xml
Printing mesh point device #8 diagnostics to mp-report-8.xml
Printing mesh point device #9 diagnostics to mp-report-9.xml
Printing mesh point device #10 diagnostics to mp-report-10.xml
Printing mesh point device #11 diagnostics to mp-report-11.xml

real	0m48.401s
user	0m48.153s
sys	0m0.240s