Re: Idle time reported by top / 386 compiling kernel

Pavel Machek (pavel@elf.ucw.cz)
Thu, 19 Mar 1998 23:03:53 +0100


Hi!

> I know that I need to upgrade my system, but I now have
> a 386DX33 (AMD & copro) and 20 Mbytes of RAM.

No, don't upgrade that beast. Slow machines result in better
software, and I do not want to be last one running 386/8Meg.

> I was compilling 2.1.90 yesterday, and I noticed (using top
> with a "nice" priority of 19) that even without swapping
> or accessing the disk (no disk light or noise!) I have
> sometime idle times of 50-70% - "cc1" and other "make" do
> not take most of the CPU time available.

Ok, I'll clean my kernel and try it... Sorry, my machine has not
enough RAM for this test. I would appredicate if you could reproduce
this bug with disk not used at all...

Hmm, yes, it looks like I can see it. But it looks like measurement
error. TOP uses idle=100%-sum{times of all processes}. Fix top.

Yes, top is lying badly. I've just seen it telling me 97% idle after
period of complete CPU load (according to debuging leds.)

I guess that this happens:

CPU is loaded by process 'A' which exits just at the end of measurement
interval.

> Is that in relation with "cc1" deacreasing its priority
> to 20 while building the kernel ?

No.

> Is that a well known problem in the sckeduler ?

No.

> Is that a measurement problem (Kernel profiler
> Define not set) ?

Kernel profile does not have anything to do with this.

Pavel

-- 
I'm really pavel@atrey.karlin.mff.cuni.cz. 	   Pavel
Look at http://atrey.karlin.mff.cuni.cz/~pavel/ ;-).

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu