Thu, 25 Dec 2008 09:18:10 -0600 (CST) Duane Attaway dattaway@dattaway.net wrote:
I ran Steven Rostedt's streamline.pl script to remove unloaded modules from the current .config.......Is there an easy way to detect unused kernel code from the running kernel and de-config it?
Actually, the question is "is there value in such an exercise on modern equipment?" I'm not sure that there isn't, but I'm far from certain that there is. I can see this dramatically cutting down build time (relevant maybe once ever 2-3 weeks), and I can see this having a substantial impact on embedded equipment (only really relevant on the first build), but neither of those things really effects what one does on a day to day basis. Given modern equipment, is there a substantial benefit to jumping through all of these extra hoops to optimize a kernel? On a related note, are there any substantial and significant reasons not to just run stock kernels at this point?
On Fri, 26 Dec 2008, Sean Crago wrote:
Actually, the question is "is there value in such an exercise on modern equipment?" I'm not sure that there isn't, but I'm far from certain that there is. I can see this dramatically cutting down build time (relevant maybe once ever 2-3 weeks), and I can see this having a substantial impact on embedded equipment (only really relevant on the first build), but neither of those things really effects what one does on a day to day basis. Given modern equipment, is there a substantial benefit to jumping through all of these extra hoops to optimize a kernel? On a related note, are there any substantial and significant reasons not to just run stock kernels at this point?
I have a few ARM powered boards with 32MB and less. Small kernel size makes a big difference in embedded equipment. Sometimes the memory and cache size is kept small, which greatly reduces power. An optimized kernel makes a big difference.
-=Duane http://www.dattaway.net
-
Duane Attaway -
Sean Crago