On Fri, Sep 5, 2003 at 11:09pm, Christopher SEKIYA wrote:
> Subject: New-and-improved snapshot available

> All,
> 
> Thanks to feedback received about my previous snapshot, I've made another
> available at ftp://ftp.rezrov.net/pub/NetBSD-sgimips/snapshot-20030904
> 
> Changes since the last snapshot are:
> 
> * real r10k support, verified to be 99.(44/100) working.
> * no pretense of an ARCS console on ip32.
> * more tweaks to the mace interrupt code.  com2 access will still fry the
>   system, as there's an interrupt masking problem that I haven't figured out.

OK, Chris... This snapshot is tonnes better'n what I was working with
previously on my R10K O2.  

The mace CPU error has gone away.  
The machine is MUCH more stable.  No crashes yet, and I'm well into 
	compiling the commands in a full build (previously, I wouldn't
	make it past the first part of the libraries).
The machine is VERY MUCH noticeably faster somehow.  I'd say double
	the speed I was getting before.  I don't know if this is
	a cache thing, or if somehow SCSI is working faster.  Since
	the drive chatters away constantly, I'm guessing there's
	somewhat an I/O bounding happening with a build.
	Regard:  4:29PM  up  7:30, 2 users, load averages: 1.57, 1.80, 1.90
	Previously, the 15-minute average was well into the 4's
	constantly.

I have yet to test a kernel-INSTALL image yet, as I'm anxious to determine
the stability of the current kernel.  When the build finishes I'll be doing
that as well as generating a bootable-CD install image and moving onwards
with creating that image via a "release" build.  Perhaps I'll even ponder
that "cross-platform" bit we already discussed ;)

Um... Actually... Now that I think of it, I might as well do test installs
on the INDYs, INDIGO 2 variants and an R5K O2.

Enjoying the cookie dough? :D

-scott