I constantly upgrade my kernel and constantly keep applications updated, I have been reading many sites proclaiming that CFS patchset "The Completely Fair Scheduler" from Ingo Molnar is better and faster than Con Kolivas 2.6.22-ck1 patchset
I am extremely dissapointed in all the hype and the news that cfs will be a part of kernel 2.6.23...
Here is what I noticed about cfs which sucks...
Boot time:5 seconds longer than ck patch
Nautilus: Takes a second longer to load up... wtf
Firefox: takes 2 more seconds to load, simple web pages like digg and my site spike to 100 cpu usage and mouse freezes..
Music: Choppy with more cpu usage
Video: glxgears reported a loss of over 100 fps
Overall CFS sucks, and ck's patchset kicks ass, I have ck's patchset installed with the latest ubuntu feisty kernel as well as the "The Completely Slow Scheduler CFS"
I would love to see some benchmarks of how cfs is faster than CK and what it actually improves. Also I cannot believe Linus is implementing this POS Patchset..
This is a true dissapointment for the Linux Community, I now understand why Con Kolivas quit because there is so many ignorant people out there that dont respect a good coders work. Ck's patchset is way faster than CFS, CFS is garbage and im only keeping it on my pc to do benchmarks when requested.
Be smart get Con Kolivas Patchset here and do you and the Linux community a favor and actually try out CFS and let me know what you experience
Here are my benchmark results with hardinfo:
cK patchset
cfs the slow scheduler
You're not doing a proper comparison. First, if you were doing these tests with an RC version of 2.6.23, you have to realize that it's not ready yet (hence the RC) and that there WILL be (and have been) updates to CFS. Second, you're comparing the ck patchset to prolly the stock kernel. You have to realize that the whole patchset has more optimizations for desktop usage than just the CPU scheduler. Also, do a real comparison...Compare a stock 2.6.22 kernel (or any stock kernel after 2.6.18) which uses the old O1 scheduler to the new 2.6.23 kernel which uses CFS as probably 90-95% of linux users are using a stock kernel.
ReplyDeletehey jeff, thankyou for the reply
ReplyDeleteI have benchmark results with hardinfo showing that ck's patchset is more responsive for desktop use vs CFS, ck beat CFS in each benchmark and the results are very noticeable in a desktop environment. I will be doing a comparison of the generic ubuntu kernel's and maybe a stock kernel from kernel.org.
Did you post your findings to the linux mailing list? If not, please do so, then Ingo can work on improving the scheduler for that workload. Just sitting there and complaining does not help anyone.
ReplyDelete