dead code removal

dead code removal

[Jim Reekes]

I have a little program that contains a routine DummyFunc() which is
never reference in the code. Building this single source file in any way
I've tried results in a binary which contains DummyFunc(). Using nm and
objdump, as far as I can determine GCC 2.95 does not dead-code strip.
(which I find amazing!). So I have two questions.

Am I missing something and GCC actually does dead-code stripping?

 - or - 

How can I find unused/unreferenced symbols in my code so that I can bury
the dead code manually?


Jim

P.S. I have always compared programming to ditch digging, and now I'm
convinced.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: dead code removal

[Daniel Jacobowitz]

That kind of stripping can not be done by the compiler.  It could
theoretically be done by the linker, and there MIGHT be a linker option
to give the warnings you want - what if another object file in your
program called DummyFunc?  Or worse, what if a shared library expected
DummyFunc to be available?


On Wed, Nov 24, 1999 at 05:10:03PM -0800, Jim Reekes wrote:
> 
> I have a little program that contains a routine DummyFunc() which is
> never reference in the code. Building this single source file in any way
> I've tried results in a binary which contains DummyFunc(). Using nm and
> objdump, as far as I can determine GCC 2.95 does not dead-code strip.
> (which I find amazing!). So I have two questions.
> 
> Am I missing something and GCC actually does dead-code stripping?
> 
>  - or - 
> 
> How can I find unused/unreferenced symbols in my code so that I can bury
> the dead code manually?
> 
> 
> Jim
> 
> P.S. I have always compared programming to ditch digging, and now I'm
> convinced.
> 


Dan

/--------------------------------\  /--------------------------------\
|       Daniel Jacobowitz        |__|        SCS Class of 2002       |
|   Debian GNU/Linux Developer    __    Carnegie Mellon University   |
|         dan@debian.org         |  |       dmj+@andrew.cmu.edu      |
\--------------------------------/  \--------------------------------/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: dead code removal

[Jeffrey Karl Lassahn]

On Thu, 25 Nov 1999, Daniel Jacobowitz wrote:

> 
> That kind of stripping can not be done by the compiler.  It could
> theoretically be done by the linker, and there MIGHT be a linker option
> to give the warnings you want - what if another object file in your
> program called DummyFunc?  Or worse, what if a shared library expected
> DummyFunc to be available?
>
In principle it can be if the function is marked static (so the compiler
knows it will never link outside the module.)  Does anyone know if GCC
does remove the code in this case?

 
> 
> On Wed, Nov 24, 1999 at 05:10:03PM -0800, Jim Reekes wrote:
> > 
> > I have a little program that contains a routine DummyFunc() which is
> > never reference in the code. Building this single source file in any way
> > I've tried results in a binary which contains DummyFunc(). Using nm and
> > objdump, as far as I can determine GCC 2.95 does not dead-code strip.
> > (which I find amazing!). So I have two questions.
> > 
> > Am I missing something and GCC actually does dead-code stripping?
> > 
> >  - or - 
> > 
> > How can I find unused/unreferenced symbols in my code so that I can bury
> > the dead code manually?
> > 
> > 
> > Jim
> > 
> > P.S. I have always compared programming to ditch digging, and now I'm
> > convinced.
> > 
> 
> 
> Dan
> 
> /--------------------------------\  /--------------------------------\
> |       Daniel Jacobowitz        |__|        SCS Class of 2002       |
> |   Debian GNU/Linux Developer    __    Carnegie Mellon University   |
> |         dan@debian.org         |  |       dmj+@andrew.cmu.edu      |
> \--------------------------------/  \--------------------------------/
> 

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: dead code removal

[Jeremy Fitzhardinge]

On 25-Nov-99 Jeffrey Karl Lassahn wrote:
> In principle it can be if the function is marked static (so the compiler
> knows it will never link outside the module.)  Does anyone know if GCC
> does remove the code in this case?

No, only if its inline.  Just because the compiler can't see a reference, it
doesn't mean one isn't there - inline assember may refer to the static function.

Basically, you need the linker to strip dead functions, because nothing else
has enough information to do it right.

        J

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/