[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE: simpler lockobj for back-bdb entry cache locking

Another combination.

No local malloc + simple lockobj + without (objectClass=referral) in

servers/slapd/slapd : 46.1%
libc : 20.48%
libpthread : 13.1%
vmlinux : 12.7%
libdb : 1.6%


DirMark : around 2450 ops/sec ( lags the local malloc case by 50~100 )

So lots of libc and libpthread overhead has been reduced by the local
malloc patch.
This will pay off more as we optimize another part such as threading and