This is the mail archive of the gcc-patches@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: libtool (was Re: [patch] releases.html)


On Thu, Feb 15, 2001 at 06:29:41PM -0200, Alexandre Oliva wrote:
> 
> You're welcome to play with the AIX 4.1 box with 16 MB of memory I've
> got to play with.  I'm amazed every time it bootstraps a snapshot
> before the following snapshot is released :-)

Well, you've now given us a new measurement for how slow is "too slow" --
when bootstrapping one snapshot runs into the following snapshot, we know
we've hit rock bottom.  :-)


> > I'm already contemplating ways to get Make to run cc1 directly - heh,
> > now there's where we may want dynamic module loading, come to think of
> > it.  The sanest approach I can think of involves making a dynamic
> > module for Make to pull in when it notices that the compiler supports
> > this.
> 
> This would be *way* cool.  make could fork N (from -jN) cc1
> processes/threads/whatever at start-up, and keep telling them what to
> do through pipes.  No start-up cost at all!

I wish I could find that paper where some researchers wrote their linker
to do its normal work on the first invocation, then fork off and run
as a daemon.  Future relinks contacted the linker daemon.  Since there
was a symbol table and whatnot already constructed, link time dropped
*significantly*.  I've often wondered what that would be like for GCC and
compilation time; a little gccd running during the length of a project
build.

Boy, if only gld and gas supported more platforms.  Just imagine all of the
Nifty Things[tm] we could do if we required binutils to be present rather
than relying on the native tools.  (Of course, if we did that, we would
ship binutils along with gcc.)  We could share the same symbol table from
the preprocessor all the way to the linker, for example.  But until then,
that's all we can do:  just imagining.  :-)


Phil

-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.


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