Repository navigation
Aggregate core usage on many-core systems #999
Description
On massive multi-core systems (more than 128 cores), even the aggregation on 4 columns makes it impossible to view effectively the process panel as the full space is used for displaying the core usage.
One way to mitigate that is to restrict the view to Average CPU (F2 - delete "CPU (1&2)/4" and "CPU (3&4)/4") translating in the following change in ~/.config/htop/htoprc:
account_guest_in_cpu_meter=0
color_scheme=0
delay=15
-left_meters=LeftCPUs2 CPU Memory Swap
-left_meter_modes=1 1 1 1
-right_meters=RightCPUs2 Tasks LoadAverage Uptime
-right_meter_modes=1 2 2 2
+left_meters=CPU Memory Swap
+left_meter_modes=1 1 1
+right_meters=Tasks LoadAverage Uptime
+right_meter_modes=2 2 2Yet this average view is of less interest for debugging parallel jobs / processes and quickly review tasks affinity etc: there should be an intermediate solution between the detailed core view and the aggregation on a single bar.
So would it be possible to offer more options for aggregation of the meter (currently at max 4 columns-- could we imagine for instance going on 8 columns and/or aggregate statistics of 2 cores per line for instance ?
Activity
antoinecarme commentedon May 26, 2020
htop rocks. Thanks
I have a similar issue with a 64 cores / 256 threads system ...
see
https://github.com/antoinecarme/xeon-phi-data/blob/master/asrock/htop-fail.png
When setting a tiny font in the terminal, I get something like this
https://github.com/antoinecarme/xeon-phi-data/blob/master/asrock/htop-xeon-phi-tiny-font.png
Full screen is just enough to display everything. An aggregated CPU meter column view is welcome.
I know that it is not easy to adapt htop to this kind of system (6 columns instead of 4 ?).