On 16.07.05, 12:46:52, Michael wrote:
> 
> This seems to be a generic problem - people with all kinds of sparc
> hardware complained about processes dying seemingly at random.

Reassuring that it's not my hardware that has gone faulty :)

> 
> That would be helpful - I didn't see a kernel panic that wasn't related
> to me hacking something for ages. As far as I know the only difference
> between the 3GX and the 3TX is the CPU ( and therefore L2 cache ).
> Did older SPARCBOOK kernels work better? Maybe it's the CPU power saving
> stuff, I'm not sure it's doing The Right Thing(tm) for TurboSPARC CPUs.

It still appears to run pretty hot, case temperature around 38C ...
I didn't measure it with 2.0, so I wouldn't know if it's significantly
better, but it's still too warm to keep in the lap.

> 
> Which one? Xsun won't work with the SPARCBOOK kernels, mainly because
> the display driver doesn't try to emulate a cg3 anymore. The XFree86
> stuff from ftp://ftp.netbsd.org/pub/NetBSD/misc/macallan/sparc/
> should work.

Oh, I wasn't aware of that. I'll try that one, thanks.

> 
> Try the userland from
> ftp://ftp.netbsd.org/pub/NetBSD/misc/macallan/sparc/sets/ - works
> tolerably well here. No idea about releng snapshots, I always built my
> own ( mainly for -mcpu=v8 )

Even without explicit v8 optimisation, ssh appears to be a lot
faster on -current than it was on 2.0. My gut feeling says about
twice as fast, but I may be wrong.

But if it's a general problem with randomly dying processes, I
might as well stick with my own builds. I thought maybe my snapshot
was particularly flaky.

> 
> have fun
> Michael


-- 
Bernd Sieker

NetBSD: A drinking group with a serious computing problem.