This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: linux 2.1.117 built with egcs-1.1 19980823


Andi Kleen <ak@muc.de> said:
> In article
> <Pine.GSO.3.96.980823205649.9780B-100000@chateau.georgiatech-metz.fr, damon
> LOVE <dlove@georgiatech-metz.fr> writes:

> > Using egcs-2.91.54 19980823 (gcc2 ss-980609 experimental)
> > I was able to build linux kernel 2.1.117 on a redhat 5.1
> > machine... (amd k6 233)

> I did the same, but noticed a 50k increase of the image compared to
> the same tree build with gcc 2.7.2.3. I'm not sure about the reasons
> yet.

The latest egcs snapshots contain optimizations that sometimes duplicate
computations to enable other optimizations to go through. You are compiling
for different CPUs too (i486 vs i686), where alignments are different (no,
neither of the sets used is optimal AFAIK) and the best code sequences are
different too.

> 2.7.2.3 options: -O2 -m486 -fno-strength-reduce -malign-loops=2 
> 		 -malign-functions=2 -malign-jumps=2 -fomit-frame-pointer

The -fno-strength-reduce flag has no effect, it is disabled in 2.7.2.3
anyway (but not before!)

> egcs options:	-O2 -mpentiumpro -fno-strength-reduce -fomit-frame-pointer

-fno-strength-reduce isn't needed by egcs, that bug has been fixed IIRC.
--
Dr. Horst H. von Brand                       mailto:vonbrand@inf.utfsm.cl
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]