cvs commit: src/sys/i386/include endian.h

[Bernd Walter]

On Tue, Sep 23, 2003 at 09:45:31AM -0700, Peter Wemm wrote:
> Robert Watson wrote:
> > 
> > On Mon, 22 Sep 2003, Peter Wemm wrote:
> > 
> > >   Microoptimization to allow the compiler to evaluate ntohl() etc on
> > >   known constants at compile time rather than at run time.  We have a numbe
>     r
> > >   of nasty hacks around the place to cache ntohl() of constants (eg: nfs).
> > >   This change allows the compiler to compile-time evaluate ntohl(1) as
> > >   0x01000000 rather than having to emit assembler code to do it.  This
> > >   has other smaller flow-on effects because the compiler can see that
> > >   ntohl(constant) itself has a constant value now and can propagate the
> > >   compile time evaluation.
> > 
> > So should we now be doing a sweep through the tree to remove instances of
> > this?  I.e.:
> > 
> >         /*
> >          * To avoid byte-swapping the same value over and over again.
> >          */
> >         igmp_all_hosts_group = htonl(INADDR_ALLHOSTS_GROUP);
> >         igmp_all_rtrs_group = htonl(INADDR_ALLRTRS_GROUP);
> > 
> > Etc?
> 
> It is probably premature to do that just yet (only 2 of 5 or so platforms
> handled), but that was the kind of "optimization" that prompted this.
> NFS is also riddled with this.

2 platforms is already much - considered that ppc and sparc64 are big
endian and therefor not influenced.
But I think the above assignment can be optimized away by the compiler
in many cases now.
Nevertheless - having alpha done first would be fine ;-)

-- 
B.Walter                   BWCT                http://www.bwct.de
ticso at bwct.de                                  info at bwct.de