Re: high sys time, very very slow builds on new 24-core system

[Dennis Ferguson <dennis.c.ferguson%gmail.com@localhost>]

On 23 Mar 2011, at 14:24 , Thor Lancelot Simon wrote:
> Building with sources on a fast SSD ("preloaded" into the page
> cache before the build using tar > /dev/null) and obj, dest, and rel
> dirs on tmpfs, system builds are extraordinarily slow.  The system
> takes about 20 minutes to build a netbsd-5 based source tree
> with -j24 -- about the same amount of time as an older 8-core Intel
> based system running netbsd-5 requires with -j8.
> 
> All cores spend well over 50% time in 'sys', even when all or almost
> all are running cc1 processes.  The kernel is amd64 -current GENERIC
> from about 1 week ago -- no DIAGNOSTIC, DEBUG, KMEMSTATS, LOCKDEBUG,
> etc.
> 
> Does anyone have any idea what might be wrong here?

Does any of the chipset hardware show up at boot time as not being
supported?

While it is unlikely to be the same thing, I have a Dell Core i7
machine that behaves pretty much like that.  It seems to be related
to disk I/O in my case; doing an `rm -rf' of a build object
directory seems to take almost forever (it takes about 10 times
as long as the 6 year old i386 machine sitting beside it).  When
the same machine was running Linux it was exceedingly speedy.

When I boot NetBSD I get this

vendor 0x8086 product 0x342e (interrupt system, revision 0x13) at pci0 dev 20 
function 0 not configured
vendor 0x8086 product 0x3422 (interrupt system, revision 0x13) at pci0 dev 20 
function 1 not configured
vendor 0x8086 product 0x3423 (interrupt system, revision 0x13) at pci0 dev 20 
function 2 not configured
vendor 0x8086 product 0x3438 (interrupt system, revision 0x13) at pci0 dev 20 
function 3 not configured

which I was guessing might have something to do with it.

Dennis Ferguson