Ok, so the next version of the script (out soon) will indeed include single-decimal precision, just like your example of 45.6% :)

But for anyone who can't wait for the next version, the added decimal is very easy to do... just remove these two lines:

  let "DIFF_USAGE=(1000*($DIFF_TOTAL-$DIFF_IDLE)/$DIFF_TOTAL+5)/10"
  echo -en "\rCPU: $DIFF_USAGE%  \b\b"

and replace them with:

  let "DIFF_USAGE=1000*($DIFF_TOTAL-$DIFF_IDLE)/$DIFF_TOTAL"
  let "DIFF_USAGE_UNITS=$DIFF_USAGE/10"
  let "DIFF_USAGE_DECIMAL=$DIFF_USAGE%10"
  echo -en "\rCPU: $DIFF_USAGE_UNITS.$DIFF_USAGE_DECIMAL%    \b\b\b\b"

Of course, the same approach can be extended to multiple decimal places, but anything more than a single decimal place would be implying a false sense of accuracy, I would think.

Anyway stay tuned for the official 1.1 version soon... ;)

That's pretty cool, too bad he didn't link back to you. :-(

Oh well... Still nice to see someone utilizing it... and your name is still in there!!!

It should not be too hard to do what you have in mind. Basically, you'd need to do two things to the above script for it work:

1. strip the "while" loop out completely.

2. save/restore PREV_TOTAL and PREV_IDLE via temporary files.

Since the script depends on knowing the previous total and idle times (the only way to calculate the "differences" needed to find the "current" CPU load), those times will need to be stored somewhere.

So, at the start of the script, check for the existence of the temp files eg /tmp/prevCpuIdle.txt and /tmp/prevCpuTotal.txt, and initialise PREV_TOTAL and PREV_IDLE to the contents of those files accordingly, or 0 if they don't exist yet.

Then, at the end of the script, simply write out the $TOTAL and $IDLE times to those two files.

Of course, there's many other things to consider if you want it to be robust (eg how often does the machine wipe /tmp, and will this script *ever* run multiple instances simultaneously). But that should get you going :)

PS - I'd probably write the temp files to /shm or /dev/shm if you have one - especially if you will be running this script often! ;)

paul.

I would never have thought to use it in a $PS1... that's a brilliant idea!! I'll definitely give that ago sometime soon :)

Looks doable. You'd want to get the utime and stime (and maybe cutime and cstime, if you want to include children). But they're in clock ticks, which you'd need to compare to the totals from /proc/stat, which are in USER_HZ... would take a little bit of investigating to figure out a reliable cross-platform way to compare clock ticks with USER_HZ (it's probably really easy, but I've just never thought about it).

Hmm... will have to look deeper another time - you've got my curiosity up now ;)

pc.

Yeah, I'm working on a version that supports either mode :)

But in the meantime, check out my comment above for a quick explanation of how you can do it yourself if you like ;)

pc.

Hi Brendon, that's looks pretty cool! :)

Thanks for letting me know, and for including a reference to this blog post in the README file!! :)

pc.

Thanks for sharing that back on this blog. I might never have noticed it.
I'll load it up on one of my servers soon and see how it goes.

Great job.

-Mike

I don't know much about other shells... what shell are you using?

The error message suggests to me that the error is somewhere *before* line 10... have you made any changes to the script above that line? (Since the quoted line is #8 in my script, I suppose you probably have).

Feel free to paste the script as you have it now, and I'll take a look :)

pc.

I got the same error when exec this script,but if you use this command "/bin/bash script_name.sh",It will be ok.....
I think ,This script just not compatible with sh....

You're very welcome! Thanks for letting me know :)

pc.

Thanks, you're quite right! :)

I guess, depending on the desired result, one could either use something like 'head' as you suggested to select just the combined-CPUs line (perhaps with an optional script argument to specify which CPU to monitor instead), or track each CPU individually and print them all as individual percentages.

I might implement one, or both sometime... :)

Thanks again!

paul.

That hack is funny and clever.
I love the idea of a 3 hour hack, but I'm pretty sure my 3 hours would end up being spread out. I also think I'd end up spending 5 hours because that's how I roll. lol.