Re: [PATCH] backing dev unplugging

From: David Mosberger
Date: Wed Mar 10 2004 - 19:08:03 EST

>>>>> On Wed, 10 Mar 2004 15:54:19 -0800, Andrew Morton <akpm@xxxxxxxx> said:

Andrew> jbarnes@xxxxxxx (Jesse Barnes) wrote:
>> -------------------------------------
>> w/Jens' patch: ~47149 I/Os per second

Andrew> Happier.

>> [root@revenue sio]# readprofile -m /root/ | sort -nr +2
>> | head -20 181993 default_idle 5687.2812 624772 snidle 1627.0104
>> 209129 cpu_idle 435.6854 4755 dio_bio_end_io 12.3828 6593
>> scsi_end_request 12.1195 435 ia64_spinlock_contention 6.7969 2959
>> sn_dma_flush 4.4033

Andrew> Do you know where that spinlock contention is coming from?

Andrew> (We have a little patch for x86 which places the spinning
Andrew> code inline in the caller of spin_lock() so it appears
Andrew> nicely in profiles.)

And real men use profiling tools that provide a call-graph, so this
hack isn't necessary... ;-)

Jesse, if you want to try q-tools and need some help in getting the
per-CPU results merged, let me know (it's something that's planned for
a future release, but there is only so many hours in a day so this
hasn't been done yet).

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at