You are not logged in.
vicious graph widgets in awesome 3.5 do need more cpu cycles than with awesome 3.4.
I do have a cpuwidged declared in rc.lua:
this is the same in 3.4 and 3.5
cpuwidget = awful.widget.graph()
vicious.register(cpuwidget, vicious.widgets.cpu, "$1")additionally in awesome 3.4 rc.lua:
mywibox[s].widgets = {
[...]
cpuwidget.widget,
[...]
}additionally in awesome 3.5 rc.lua:
local right_layout = wibox.layout.fixed.horizontal()
[...]
right_layout:add(cpuwidget)(... 3.5 lua code changes ... you know ...)
Viewing "top" with Awesome 3.4 -> the awesome process needs below 0.5 %CPU load if in the top list at all.
Viewing "top" with Awesome 3.5 -> the awesome process needs ca. 3 %CPU.
I didn't tested it, but I'm sure powertop will show much more wakeups with 3.5 than with 3.4.
I removed the cpuwidget in 3.5 and it also remains at <1%. I did add a second graph and the cpu load climbs even higher than 5%.
I've found a simple bash script for collecting CPU Load values. This way there are many process forks (bash itself, /usr/bin/sleep, etc.) but this bash scripting solution is consuming even less ressources than awesome(3.5)+vicious.
In my eyes this is a big regression. I've downgraded to Awesome 3.4 for now. Observing development...
Last edited by hede (2013-01-03 12:25:18)
Offline
I'm still using awesome 3.4 here because of the higher CPU load with awesome 3.5.
If I do not register any vicious widget with awesome but running graph updates via awesome-client then 3.4 is fine where 3.5 has higher CPU load / consumes more CPU cycles.
i.e. in rc.lua:
cpuwidget = awful.widget.graph()
-- vicious.register(cpuwidget, vicious.widgets.cpu, "$1", 2)
cpuwidget:add_value(0.5)using a shell:
while /bin/true; do echo "cpuwidget:add_value(0.5)" | awesome-client ; sleep 3; doneresults in ~0% CPU load with awesome 3.4 and ~3% CPU load with awesome 3.5. (1.6 GHz Intel Core here)
So there's still some regression with awesome 3.5.1-1 and it seems vicious is not the reason.
Offline