Jim Fiori wrote:
Lock contention Matt. See what locks are hot:
# lockstat -o lockstat.out -D20 sleep 10
Matt V. wrote:
OK, I tried the DTrace script "hotkernel" et the result is:
[...]
ce`ce_intr 1245 0.4%
unix`page_lookup_nowait 1330 0.5%
genunix`rm_assize 1448 0.5%
genunix`segmap_getmapflt 1578 0.5%
genunix`vmu_calculate_seg 1698 0.6%
SUNW,UltraSPARC-IV`gettick 1823 0.6%
unix`bzero 2157 0.7%
unix`page_lookup_create 2209 0.8%
SUNW,UltraSPARC-IV`copyin_more 2269 0.8%
genunix`avl_walk 2286 0.8%
SUNW,UltraSPARC-IV`send_mondo_set 2309 0.8%
SUNW,UltraSPARC-IV`copyout_more 2530 0.9%
SUNW,UltraSPARC-IV`bcopy 2564 0.9%
unix`page_trylock 2711 0.9%
unix`mutex_owner_running 3363 1.2%
unix`page_exists 6052 2.1%
unix`idle 6709 2.3%
unix`default_lock_delay 9016 3.1%
unix`mutex_enter 11670 4.0%
unix`disp_getwork 56868 19.6%
unix`mutex_delay_default 96390 33.2%
Now, I know that is 'mutex_delay_default' and 'disp_getwork' which
work a lot but which process or PID is behind please ?
disp_getwork() isn't necessarily interesting here..as that routine is
basically called out of the idle CPU logic (so if CPU's are idle,
there's a good chance they will be trolling around for work in
disp_getwork()). The mutex stuff is interesting though ;)
-Eric
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
