The log says:
# cat /tmp/warn.log
0: Detected 160 online CPU.
0: calibrate_cpuinfo: nsofclk = 0.4167, clkofsec = 2400000000
5: perf_init() is failed
The kernel crashed shortly after I saw this. But, either way, it would
be nice for numatop to give _some_ indication on the console of why it
Also, the "-l 2" option for setting the logging level is a bit unusual.
Most programs use "-dX" for debug levels.
I was running some kernel compiles and ran numatop while they were in
progress. I noticed a few things.
A lot of the compiles were short-lived. If they lived less than the
sampling frequency, numatop might never display them. One suggestion
would be to collate the entire system worth of stats (those for every
task) and display RMA/LMA at the top for the _whole_ system, just like
the real 'top' does for CPU.
You might also want to try running numatop while on/offlining cpus. For
me, it just fell over and started displaying garbage.
Page up/down don't seem to do anything. Should they?
Is there a way to reverse the sorting? I would like to be able to see
both the tasks with the _best_ latency ratios and the worst.