[yocto] Build performance with Toaster analysis

Damian, Alexandru alexandru.damian at intel.com
Wed Nov 13 06:30:55 PST 2013


Hello,

Just to make sure that Toaster doesn't have a big performance impact, I've
ran a smoke test for a build with and without Toaster enabled.

For the test, I've ran a core-image-build with the default out-of-box
configuration on origin/dora-toaster, 4 parallel jobs and -j 8 for make.
Both builds were done on clean directories, on the same filesystem, after a
clean reboot, with no other tasks running on the machine. Timing was
measured with the time utility. A single build was ran for each
configuration. The comparison for the wall time is:

- without Toaster: 63m35
- with Toaster: 62m21

The difference is less than 2%, which I attribute to the noise in the
system, and shows that Toaster has no significant impact in build
performance at this point. To determine the true impact of Toaster, a
number of builds for each configuration needs to be run, and get the median
and average timing for each mode in order to get a reasonable conclusion.

Cheers,
Alex
-- 
Alex Damian
Yocto Project
SSG / OTC
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.yoctoproject.org/pipermail/yocto/attachments/20131113/a6999656/attachment.html>


More information about the yocto mailing list