RE: boot-time slowdown for measure_migration_cost

From: Luck, Tony
Date: Fri Jan 27 2006 - 16:46:51 EST


> The boot-time migration cost auto-tuning stuff seems to have
> been merged to Linus' tree since 2.6.15. On little one- or
> two-processor systems, the time required to measure the
> migration costs isn't very noticeable, but by the time we
> get to even a four-processor ia64 box, it adds about
> 30 seconds to the boot time, which seems like a lot.

I only see about 16 seconds for a 4-way tiger (not that 16 seconds
is good ... but it not as bad as 30). This was with a build
from tiger_defconfig that sets CONFIG_NR_CPUS=4 ... so I wonder
what's causing the factor of two. I measured with a printk
each side of build_sched_domains() and booted with the "time"
command line arg to get:

[ 0.540718] Building sched domains
[ 16.124693] migration_cost=10091
[ 16.124789] Done

More importantly, how does this time scale as the number of
cpus increases? Linear, or worse? What happens on a 512 cpu
Altix (if it's quadratic, they may be still waiting for the
boot to finish :-)

-Tony
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/