CPU usage on Linux

pepi37 · 2022-03-13, 21:38

I wrote small script to show me cpu usage, temperature of CPU , used memory, uptime...

what is my problem?

When I run mprime under linux then script show some percentage of usage (nearly every time 100%)

since it run on all cores

Quote:

echo
top -n 1 | awk 'NR>7 && $2=="root" { sum += $9 } END { print "CPU USAGE:" (sum/800)*100" %" }'
echo

Currently I run boinc clinet with 8 LLR2 task and CPU usage is zero?

Quote:

RYZEN 3700X
CURRENT DATE AND TIME
13-03-2022 22:36:38
TOTAL MEMORY = 7.78075 GB
CPU TEMPERATURE
+49.8°C
***** UP TIME *****
2 weeks, 3 days, 12 hours, 31 minute
CPU USAGE:0 %

Where is error in my script?

paulunderwood · 2022-03-13, 23:00

Try sum*100/800 or simply sum/8.

pepi37 · 2022-03-14, 06:50

Quote:

Originally Posted by paulunderwood

Try sum*100/800 or simply sum/8.

Thanks, I will try it, but I asked why is cpu usage process dependent? When run mprime show correct value, but when I run llr2 and Boinc show zero usage?

retina · 2022-03-14, 07:33

Quote:

Originally Posted by pepi37

Thanks, I will try it, but I asked why is cpu usage process dependent? When run mprime show correct value, but when I run llr2 and Boinc show zero usage?

Maybe those other programs don't run as root?

retina · 2022-03-14, 22:20

Perhaps a more robust summation script would be this:

CORES=$(grep stepping /proc/cpuinfo | wc -l) ; awk "{ total+=\$1 } END { print total/$CORES }" <(ps -u $USER -o %cpu h)

If you wanted track just for particular programs then maybe this:

CORES=$(grep stepping /proc/cpuinfo | wc -l) ; awk "{ total+=\$1 } END { print total/$CORES }" <(ps -C mprime,boinc -o %cpu h)

Or for all processes:

CORES=$(grep stepping /proc/cpuinfo | wc -l) ; awk "{ total+=\$1 } END { print total/$CORES }" <(ps -o %cpu h)

chris2be8 · 2022-03-15, 16:46

Or:
CORES=$(grep -c stepping /proc/cpuinfo)

Which is slightly faster, but I'd only run it once at the start of a script so speed wouldn't make much difference.

Mark Rose · 2022-04-21, 06:03

Or CORES=$(nproc)

retina · 2022-04-21, 06:08

Quote:

Originally Posted by Mark Rose

Or CORES=$(nproc)

Yeah, better.

But also nproc --all to ensure you count all cores, not just those in the current process.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Windows Subsystem for Linux v2 gets real Linux kernel	tServo	Software	0	2019-05-07 16:59
GPU Usage	Brain	GPU Computing	9	2011-04-12 22:25
Usage of GMP-ECM	ECMFreak	Factoring	13	2007-07-20 17:34
max heap usage of Linux process	prasanta	Programming	1	2006-11-13 22:49
CPU usage	Unregistered	Software	6	2003-11-19 07:05

2022-03-13, 23:00	#2
paulunderwood Sep 2002 Database er0rr 3×5×281 Posts	Try sum*100/800 or simply sum/8.

2022-03-14, 22:20	#5
retina Undefined "The unspeakable one" Jun 2006 My evil lair 5·1,303 Posts	Perhaps a more robust summation script would be this: CORES=$(grep stepping /proc/cpuinfo \| wc -l) ; awk "{ total+=\$1 } END { print total/$CORES }" <(ps -u $USER -o %cpu h) If you wanted track just for particular programs then maybe this: CORES=$(grep stepping /proc/cpuinfo \| wc -l) ; awk "{ total+=\$1 } END { print total/$CORES }" <(ps -C mprime,boinc -o %cpu h) Or for all processes: CORES=$(grep stepping /proc/cpuinfo \| wc -l) ; awk "{ total+=\$1 } END { print total/$CORES }" <(ps -o %cpu h)

2022-03-15, 16:46	#6
chris2be8 Sep 2009 3⁴×29 Posts	Or: CORES=$(grep -c stepping /proc/cpuinfo) Which is slightly faster, but I'd only run it once at the start of a script so speed wouldn't make much difference.

2022-04-21, 06:03	#7
Mark Rose "/X\(‘-‘)/X\" Jan 2013 B89₁₆ Posts	Or CORES=$(nproc)