Re: High interrupt CPU usage in top

[Mark Cullen <mark.r.cullen@xxxxxxxxx>]

Matthew Dillon wrote:
> :It does boot with DMA turned on, unless I am missing something? The 
> :WDMA2 drive is the old CD drive, the UDMA33 drive is the hard disk. :-)
>
>     Yes, that is so... but I'm not entirely sure that it actually uses
>     those settings when you have a configured slave without a master.
>
>     The dd command you ran isn't really a good test.  Because you didn't
>     specify a block size, dd will use 512 bytes.  Needless to say 512 byte
>     transfers from the hard drive are going to be cpu-bound and most of the
>     overhead is going to be in command setup and tear down.  Try 32K
>     transfers:
>
>     dd if=/dev/adX of=/dev/null bs=32k
>
>     It should be < 1%.
>
> 					-Matt
> 					Matthew Dillon 
> 					<dillon@xxxxxxxxxxxxx>

I did 16k aswell, and yes it was much lower on both OS's. I figured you 
wanted a 'worst-case' sorta thing is all :P Oh, and the whole mess with 
the hard disk set up was very temporary, I was just too lazy to mess 
about changing jumpers around really.

Is the interrupt issue some sort of accounting stuff in the kernel like 
Joerg said? If so, is that debugging info of sorts which may possibly be 
removed / optional in future releases?