power measurements - alan-mushi/opencpn-low-energy GitHub Wiki
To know if our breakthrough were efficients in term of energy consumption, we needed to measure the improvements.
To measure the energy consumed by the application we actually needed to measure when the computer :
- is doing "nothing"
- executes OpenCPN
- executes OpenCPN with our modifications (AIS and frame timer are slow down, details)
In order to do that, we used the Yocto-watt watt-meter.
Set up diagram with true wire colors (some wires are white).
To obtain the logs that shows the measures results, we modified an example file of Yocto-watt (old_datalogger.php
), thus, we used PHP (and the library) with apache2 to execute our version of old_datalogger.php
. You can find the file here.
We used the same computer for all measures (HP EliteBook 2570p). Were plugged :
- The USB of Yocto-watt
- The ethernet wire
- The power wire
All measures have been took the same day, about the same time. Each measure is 5 minutes long. OpenCPN was already running before the measure starts (to avoid the startup "rush"). With Yocto-watt we can't have a measure below the second (this is quite annoying in our case).
The raw results are available here. We used a javascript piece of code to create the csv files (and compute the average power of the session). It's not used anymore but you can find it here.
Measured from Thibault's PC on Thibaud's PC | Measured from Thibaud's PC on Thibault's PC | |
---|---|---|
Without OpenCPN | 12.654 | 12.621 |
OpenCPN untouched | 12.831 | 12.6 |
OpenCPN modified | 12.848 | 12.815 |
Differences between the 2 laptops :
Thibaud run LXDE whereas Thibault run KDE 4 as window manager.
The difference of power consumption are explainable, the measures above didn't only look the power needed by OpenCPN but by the whole laptop. In order to get more accurate and that specifically target OpenCPN, we used PowerTop. The raw result are available in our in the measurements folders (in each sub folders) : here. We noticed that our modified version of OpenCPN (the one with timer modifications) consumed just a little bit less than the original version. If you look close at the both powertop reports, you will notice that the operating system's applications are taking the "space" left by the OpenCPN timers we have modified.