Re: lapic_at_2k interrukts eating CPU cycles

On Wed, 22 Jun 2005, Emanuel Strobl wrote:

>> Might have been John Baldwin.
>
> Hm, me and names is a weak point... ;)

Not as bad as me, I promise.  One someone introduces themselves at a 
conference, I'll literally forget their name by the end of the sentence 
where they say it.  I'll remember them, but their name will be gone really 
quickly.  Very embarrassing :-).

>> I have a feature request in to John to add statistics gathering on IPIs,
>> since he's currently reworking the interrupt paths.
>
> Thanks so much, very good explanation, I almost understood it completely 
> (I'm lacking a lot of basics). If only "The Design and Implementation of 
> the FreeBSD Operating System" was written like this...

If the D&I book were written this way, it would come in three volumes :-).

Part of the problem here is that we weren't exposing some of the internal 
behavior via monitoring interfaces, so what was a rearrangement of 
functionality in the kernel was less well represented in the monitoring. 
So I think the answer there is to improve our monitoring so it's more 
clear what's going on.

I'm currently working on enhanced versions of vmstat and some other tools 
to allow better per-cpu monitoring of memory load/allocation, and we 
should do something similar to interrupt monitoring, add IPI monitoring, 
and so on.

Robert N M Watson