Messages in this thread | | | Subject | Re: [performance bug] kernel building regression on 64 LCPUs machine | From | Shaohua Li <> | Date | Thu, 20 Jan 2011 10:12:04 +0800 |
| |
On Wed, 2011-01-19 at 22:32 +0800, Ted Ts'o wrote: > On Wed, Jan 19, 2011 at 09:55:48AM +0800, Alex,Shi wrote: > > Shaohua and I tested kernel building performance on latest kernel. and > > found it is drop about 15% on our 64 LCPUs NHM-EX machine on ext4 file > > system. > > What kerenl version were you comparing against? And what were you > using to benchmark this performance drop? I'm curious what your > workload was; I'm guesisng it was some kind of mixed read/write > workload? Were there any fsync()'s involved? 2.6.36 vs 2.6.37. this exists in 2.6.37-rc1. it's a kbuild, no a mixed read/write workload. kbuild hasn't fsync involved
Thanks, Shaohua
| |