Re: SMP per-CPU statistics under FreeBSD.

On Tue, Apr 25, 2006 at 05:31:34PM +0900, Hajimu UMEMOTO wrote:
> Hi,
> 
> >>>>> On Sun, 23 Apr 2006 01:10:00 +0000
> >>>>> Marco van Tol <[email protected]> said:
> 
> > > 2) You have the choice of reporting composite CPU data yourself or letting gkrellm
> > > calculate the composite data.  Just never call gkrellm_cpu_assign_composite_data()
> > > and gkrellm will do it for you.
> 
> The netbsd.c calls gkrellm_cpu_assign_composite_data().  As far as I
> looked the jhb's patch briefly, it seems kern.cp_time returns
> composite values.  I'm not sure which is better to call
> gkrellm_cpu_assign_composite_data() or not, but I attches the patch to
> call gkrellm_cpu_assign_composite_data().  Sorry but I don't test it
> actually even on single-CPU machine, yet.
> 
> Your patch has following line:
> 
> 	+static size_t	oid_pcpu_time_len = sizeof(oid_cp_time);
> 
> It should be:
> 
> 	+static size_t	oid_pcpu_time_len = sizeof(oid_pcpu_time);
> 
> This fix is also inclueded in my patch.

I just checked out your patch (unfortunately I don't have an SMP box at
work ;), but it doesn't display the data for the CPU's.  It does setup 2
CPU's for my box, which is the correct count, but they both stick at 0%,
even if I start a buildworld on my box.

I'll take a quick look to see if I can find why and will let you know.

Marco

-- 
Laat wat je niet laten kunt, en doe wat je niet doen kunt, en word miljonair.

_______________________________________________
Gkrellm mailing list
[email protected]
http://ninja.linux-phreak.biz/mailman/listinfo/gkrellm