Page: Previous  1, 2

Jos.v.W.
Status: Contributor
Joined: 25 Jan 2011
Posts: 57
Reply Quote
After some hours of use, the pf kernel too produces strange (random?) values:
:: Code ::

# ps axo pcpu,comm,pid,user | sort -nr | head -n 4
41643325 pppd         1043 root
41639565 gpm           973 root
76.0 einsteinbinary_  6224 boinc
75.6 einsteinbinary_  6399 boinc

# ps axo time,comm,pid,user | sort -nr | head -n 4
213503-23:34:35 gpm        973 root
213503-23:34:33 pppd      1043 root
06:55:35 einsteinbinary_  6224 boinc
03:00:58 einsteinbinary_  6399 boinc

Could this issue be related to the use BMS and running Boinc and Einstein@Home in the background?
Anyway I'll try the latest 3.7.0-1.dmz.6.

P.S. As I am staying in China I can't connect to ck-hack.blogspot.com to read Con's blog!
Back to top
Jos.v.W.
Status: Contributor
Joined: 25 Jan 2011
Posts: 57
Reply Quote
:: Quote ::

Please let me know if you still get that error.

Running 3.7.0-1.dmz.6 fourty eight hours I no longer get that error anymore.
Thanks!
Back to top
Strange cpu values for task rcu_preempt
Jos.v.W.
Status: Contributor
Joined: 25 Jan 2011
Posts: 57
Reply Quote
:: Quote ::

the 3.7.0-1.dmz.6 and newer now run CFS instead.


Since you seem to have re-anabled BFS (CONFIG_SCHED_BFS=y), running 3.7.0-6.dmz.2 again shows abnormal rcu_preempt %cpu values:

:: Code ::

ps axo pcpu,comm,pid,user | sort -nr | head -n 1
331775972 rcu_preempt    9 root
ps axo time,comm,pid,user | sort -nr | head -n 1
213503-23:34:52 rcu_preempt  9 root


I no longer trust the BFS scheduler as I don't want to spoil my calculations by random errors.
I 'm sorry if this issue can't be resolved I am going to use the Debian kernel instead.
Back to top
damentz
Status: Assistant
Joined: 09 Sep 2008
Posts: 1117
Reply Quote
Jos, I think the solution here is for me to make two kernels. One for BFS, and one for CFS. This will also make A/B testing easier, so we can do real science and see which kernel should be the best default.

In my case, I would probably make the BFS kernel the default kernel that the metapackage depends on, and have CFS available as an alternate. This will also increase build time 2x, but I think it's still workable for now.
Back to top
damentz
Status: Assistant
Joined: 09 Sep 2008
Posts: 1117
Reply Quote
3.7.0-8.dmz.1 is back on CFS. I'll reintroduce BFS as a separate package later.
Back to top
Display posts from previous:   
Page: Previous  1, 2
All times are GMT - 8 Hours