![](https://seccdn.libravatar.org/avatar/8ce4efecf1119dcbfc03135ad33fee46.jpg?s=120&d=mm&r=g)
Dorin Titiriga wrote:
I have read in linuxJournal that I can be able to optimize my kernel using this parameters:
CFLAGS='-o9 -funroll-loops -ffast-math -malign-double -mcpu=pentiumpro -march=pentiumpro -fomit-frame-pointer -fno-exceptions'
I have some problems
1. My cpu is a Atlon at 900 mhz, not a pentium. With what to replace pentiumpro ? 2. I have compiled at home the kernel ignoring -mcpu and -march, the kernel was compiled fine, with no errors, but after i have installed, it does nothing. It is broken !, so how to optimize my kernel ?
Some of these flags look distinctly dangerous. The -o9 should be -O9 and is pretty well guaranteed to cause problems (I have found that even -O3 can cause problems with very simple code I've written). The -malign-double flag will speed things up a fraction but at the risk of creating incompatibilities. For example, I once tried compiling gimp with this flag and found that it crashed when using jpeg files. The -mcpu=pentiumpro and -march=pentiumpro mean generate code optimised for a pentium pro and generate code that will only run on pentium pro or higher (and therefore even more optimised). I don't think there's any advantage in using both and they may even conflict. For an Athlon, the flag you want is -march=k6. If you configure your own kernel with xconfig (and probably menuconfig too), you get the option of specifying what kind of architecture you have and the configuration automatically gives you sensible optimisations. It is unwise to try to do more unless you really know what you are doing. JDL