[Ns-developers] Some results (Was: Static NS-3 (Was: icc and "patched" waf))

[Gustavo Carneiro]

2009/1/25 Timo Bingmann <timo.bingmann at student.kit.edu>

> Hello Gustavo and Mathieu,
>
> here are some speed timing results by simply running "time" on the program.
> On my 32bit CPU yes.
>
> Gustavo: your static patch didn't quite work, because it also linked the
> libns3.so to the programs, and that created weird duplicate module
> initialization problems. Attached is my quick hack-fix to get it to work, I
> guess you can do that more elegantly.
>
> I ran three programs: examples/wifi-ap, my new RandomVariable tester
> "samples/main-random-variable" (not in ns-3-dev) and
> scratch/three-nodes-capture-3d (also not in ns-3-dev).
> three-nodes-capture-3d runs three wireless nodes communicating at many
> different distances. The code is in my ns-3-scratch repo.
>
> Timing results with gcc, icc and with/without -fPIC are at
> http://idlebox.net/2008/ns-3-wifi/results/temp/speedtest.pdf
>
> No PIC really reduces the time on my three-nodes simulation, but shows only
> moderate speed-ups on the others. icc shows a small further speed-up in all
> cases.
>
> Some hard data to work with. Profiling will follow somewhen.


I see.  These are really substantial gains.  Too bad ICC is not free, and
too bad compiling without -fPIC makes the build complex and/or slow.

-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert