[Bf-committers] CMAKE vs SCONS vs NAN MAKE

Markus Schulz schulz at alpharesearch.de
Sun Apr 26 19:29:39 CEST 2009


I tried to make the builds as close as possible (with game engine and 
player etc)... however I'm not a build wizard and I don't know enough 
about build optimization. If you think I could make a build fast please 
reply with your hint and I will test again.

I did three test, first I build from scratch, than I build again to see 
how long it takes to find out nothing has changed and the last test was 
just a clean. Here are the commands I called and the results on my 
Ubuntu 9.04 64bit AMD 4 core with 8 GB ram:

CMAKE
debug build:
$ time make -j
real    1m11.254s
user    3m57.959s
sys    0m36.686s

$ time make -j
real    0m4.985s
user    0m7.764s
sys    0m2.352s

$ time make -j clean
real    0m0.656s
user    0m1.788s
sys    0m0.232s


release build:
$ time make -j
real    2m46.087s
user    9m41.156s
sys    0m47.227s

$ time make -j
real    0m10.254s
user    0m6.460s
sys    0m2.396s

$ time make -j clean
real    0m0.593s
user    0m1.692s
sys    0m0.256s



SCONS
debug build:
$ time python ./scons/scons.py -j 8 clean
real    2m4.794s
user    5m23.732s
sys    0m57.844s

$ time python ./scons/scons.py -j 8
real    0m34.629s
user    0m35.206s
sys    0m2.152s

$ time python ./scons/scons.py -j 8 clean
real    0m0.630s
user    0m0.360s
sys    0m0.276s*
*

release build:
$ time python ./scons/scons.py -j 8 clean
real    2m53.477s
user    9m59.833s
sys    0m56.904s

$ time python ./scons/scons.py -j 8
real    0m31.208s
user    0m30.970s
sys    0m1.196s

$ time python ./scons/scons.py -j 8 clean
real    0m0.506s
user    0m0.312s
sys    0m0.176s



NAN MAKE:
time make (-j was not working, build errors)
real    16m49.935s
user    15m17.681s
sys    1m41.222s

time make
real    0m20.538s
user    0m12.381s
sys    0m7.428s

time make clean
real    0m13.516s
user    0m8.133s
sys    0m4.836s

Regards,
Markus Schulz


More information about the Bf-committers mailing list