Switch from bioq_insert_tail() to bioqdisksort(). When the kernel is
authorMatthew Dillon <dillon@dragonflybsd.org>
Tue, 10 Jun 2008 17:20:52 +0000 (17:20 +0000)
committerMatthew Dillon <dillon@dragonflybsd.org>
Tue, 10 Jun 2008 17:20:52 +0000 (17:20 +0000)
commitec1b64f8ce85db74019d53368dbec249eab4035f
tree503ac8e8d5a690fa0c35087f1a744f1b5a2e80fd
parentb8826fbe94e65524befc794cd28f14e86b28cbc5
Switch from bioq_insert_tail() to bioqdisksort().  When the kernel is
juggling hundreds of I/O requests the on-disk controllers, which have limited
queue sizes, are unable to completely optimize the accesses.

This seems to make a pretty big difference on the twe device I tested with
when running blogbench on HAMMER, which can approach 1000 simultaniously
queued I/O's.
sys/dev/raid/aac/aacvar.h
sys/dev/raid/amr/amrvar.h
sys/dev/raid/ida/ida.c
sys/dev/raid/ips/ips_disk.c
sys/dev/raid/mlx/mlx.c
sys/dev/raid/mlx/mlx_compat.h
sys/dev/raid/twe/twe_compat.h
sys/dev/raid/twe/twevar.h