Hello, I'm using ananicy-cpp 1.2.0-1 on Cachy OS (it comes pre-installed)
I checked that the service is active and running, which is the case.
Sometimes Helium (had the same issue with Chromium in the past) takes 100% of CPU and the system becomes unstable (Gnome becomes non-reactive, so I got to wait a bit so it migh settle down). The fact is it only happens when opening a tab on Instagram (which I suspect to do shady stuff with JS since the spike of activity comes from the process link to this tab).
I checked the ananicy rules for Helium then for Doc-View and I see it applies -4 priority for niceness, I don't question this choice, but I think we could also apply the rule :
{ "cgroup": "cpu80", "CPUQuota": 80 }
So if I understand, ananicy apply cgroup per apps with their respectives rules. But I think unless you play a 3D fast-paced WebGL game, there is no reason a webbrowser should use 100% CPU continously.
Also when I checked the niceness of helium processes I see some inconsistencies with the rules :
{ "name": "helium", "type": "Doc-View" }
{ "name": "helium_crashpad_handler", "type": "BG_CPUIO" }
# Type: BackGround CPU/IO Load
# Background CPU/IO it's needed, but it must be as silent as possible
{ "type": "BG_CPUIO", "nice": 16, "ioclass": "idle", "sched": "idle" }
# Must have more CPU/IO time, but not so much as other apps
{ "type": "Image-View", "nice": -4 }
{ "type": "Doc-View", "nice": -4 }
For crashpad_handler niceness is -8 and sometimes -12 but it should be 16, also BG_CPUIO adjust niceness but not ionice.
other helium processes get -4 most of of the time and sometimes -8 for niceness.
Hello, I'm using ananicy-cpp 1.2.0-1 on Cachy OS (it comes pre-installed)
I checked that the service is active and running, which is the case.
Sometimes Helium (had the same issue with Chromium in the past) takes 100% of CPU and the system becomes unstable (Gnome becomes non-reactive, so I got to wait a bit so it migh settle down). The fact is it only happens when opening a tab on Instagram (which I suspect to do shady stuff with JS since the spike of activity comes from the process link to this tab).
I checked the ananicy rules for Helium then for Doc-View and I see it applies -4 priority for niceness, I don't question this choice, but I think we could also apply the rule :
{ "cgroup": "cpu80", "CPUQuota": 80 }So if I understand, ananicy apply cgroup per apps with their respectives rules. But I think unless you play a 3D fast-paced WebGL game, there is no reason a webbrowser should use 100% CPU continously.
Also when I checked the niceness of helium processes I see some inconsistencies with the rules :
For crashpad_handler niceness is -8 and sometimes -12 but it should be 16, also BG_CPUIO adjust niceness but not ionice.
other helium processes get -4 most of of the time and sometimes -8 for niceness.