Hi,

As per review mail thread, I tested the impact of enabling all data
tracking in order to gather complete history for all running functions.

The raw results are below. My conclusion is that we face a roughly 5% time
hit during command execution if we enable data history tracking at all
times.

I'm now investigating if we can minimize this impact by selectively
enabling variable history.

Raw data below. This is the averaged time in seconds over 5 runs of the
same command. Different (small) workloads were used to asses how variable
history impacts running time in various scenarios. Timed using the `time `
command, measurement of the wall clock.

We see a roughly 5% performance degradation for full data tracking enabling.

-- no op empty run core-image-minimal
no data tracking: 5.9
ya data tracking: 6.3

-- 1 task clean
no data tracking: 2.6
ya data tracking: 2.8

-- 3 task cleanall
no data tracking: 3.0
ya data tracking: 3.1

-- 14 task linux-libc-headers
no data tracking: 30.6
ya data tracking: 32.2

-- 
Alex Damian
Yocto Project
SSG / OTC
_______________________________________________
toaster mailing list
[email protected]
https://lists.yoctoproject.org/listinfo/toaster

Reply via email to