hi!
i did play with it a bit but left scons aside.
after a few tests i kicked out gnumake aswell as it used up all memory
on my machine and started swapping...
also i noticed that dmake 4.5 and dmake 4.6 make a huge difference.
running run_test.sh on a local disk:
complete hot
touch.sh 15m30s -
gnumake 132m45s 1m45s
jam 22m42s 0m12s
dmake 4.5 109m54s 1m47s
dmake 4.6 7m52s 1m45s
testing the same thing on a samba volume was rather ugly:
complete hot
touch.sh 26m16s -
jam 52m18s 2m34s
dmake 4.5 121m42s 8m58s
dmake 4.6 22m31s 8m18s
running this test with more than one process gave slight improvements
even on a machine with a single cpu.
tschau...
ause
ps: PIII 1800, 512mb
Kai Backman wrote:
> In the presentation last week there was some measurements on the raw
> speed of different build tools. Here is the test harness used:
>
> http://wiki.services.openoffice.org/mwiki/images/1/11/Buildtest.tgz
>
> Usage:
> - unpack and cd into
> - python createfiles.py
> - ./run_tests.sh
>
> It basically creates a 15k file fake project (OO.o has about 23k C++
> files), a sufficient number of dependencies and then "builds" that
> project using different tools. The update action is a simple "touch"
> and a shell script, touch.sh, is generated to measure the baseline.
> The resulting data is output into a runlog_* file.
>
> Please post results you get on various systems here, I'll collect
> them into the wiki.
>
> Thanks in advance for working with this! :-)
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
