Re: [offtopic?] process cpu load % wrong?

Tuukka Toivonen (tuukkat@stekt.oulu.fi)
Fri, 17 Oct 1997 09:19:15 +0300 (EET DST)


On Thu, 16 Oct 1997, Ingo Molnar wrote:

>if you have IDE disks, PIO mode, then this is natural. Always the current

My disk should be in multiword mode2 busmaster dma mode. But, the main
source of extra load seems to be dosemu. I guess DOS-programs cause a LOT
of exceptions which are interpreted by dosemu (like reading and writing
device i/o ports)

>process handles interrupts, no matter who requested that interrupt. It
>will be accounted to the handling process. So if you have a process that
[...]
>since interrupts cannot easily be 'personalized' (externally triggered
>interrupts have no 'owners', and there are interrupts that have many
>'owners', etc.), there is no easy 'solution'.

Couldn't an interrupt, which is caused by an unknown process, be added eg.
init-process load, or not added to any process load?

It is severely misleading when they are added to a random process load: the
percent numbers displayed by eg. top are correct only by luck. They may as
well be wrong.

| Tuukka Toivonen <tuukkat@stekt.oulu.fi>
| Homepage: http://stekt.oulu.fi/~tuukkat/
| Try also finger -l tuukkat@stekt.oulu.fi
| Studying information engineering at the University of Oulu
+-----------------------------------------------------------