TITLE:		Compiler Optimization
LFS VERSION:	2.3.5
AUTHOR:		Per Liden <per@fukt.hk-r.se>

SYNOPSIS:
		How to optimize the compilation for a certain CPU.

HINT:
To optimize for a specific architecture you can add the -m<cpu>
-march=<cpu> options when compiling. If you want to optimize for, say PII,
then you could do someting like this:

before running ./configure, type:

(if you use bash)

        export CFLAGS="-02 -mpentiumpro -march=pentiumpro"
        export CXXFLAGS="-02 -mpentiumpro -march=pentiumpro"

(or if you use tcsh)

        setenv CFLAGS "-02 -mpentiumpro -march=pentiumpro"
        setenv CXXFLAGS "-02 -mpentiumpro -march=pentiumpro"

the -O2 option could be set to something higher than 2, but 2 is a safe
choice. If the package in question doesn't have a ./configure script you
sometimes (depending on how the Makefile is written) have to edit the
Makefile by hand.

Note that when using the -march=<cpu> option, the compiler will make use
of architecture specific instructions, and therefor the binary will not
run on older architectures, in this case pentium, i486, etc.

The -m<cpu> option does not cause the compiler to generate archetecture
specific instructions, however, the compiler schedules the code to perform
best on the specified cpu.

Valid cpu (Intel and compatables) types in gcc 2.95.2 are:
486, pentium, pentiumpro, k6

/pli