Re: binary sizes

[Andrew Haley <aph at redhat dot com>]

Adam Megacz writes:
 > 
 > Andrew Haley <aph@redhat.com> writes:
 > >  > right?  Is it really supposed to consume 20% of the binary?
 > > Yes.
 > 
 > Ugh.
 > 
 > 
 > > Perhaps, but we're moving to removing the traditional stack chain by
 > > compiling with omit-frame-pointer: once that move is complete, DWARF2
 > > unwinder data will be needed everywhere.
 > 
 > Hrm, this sounds interesting, but I thought that fomit-frame-pointer
 > only drops the frame pointer for functions that don't need it.

That's right, but most functions don't need it.  You only really need
a frame pointer for things like alloca().

 > Do you mean that DWARF2 info is needed in order to use a debugger
 > or get stack traces?

Yes.

 > I can live without those...

Java can't.  There is a bunch of Java features that needs stack
traces.

 > > Some of the reflection data is used by the gc.
 > 
 > The GC?  That's wierd... what for?  Determining pointerfree regions?

It depends how you have the gc set up.  If you use the bitmap scheme
throughout, you only need the bitmap desciptors, and these are
produced by the compiler.

Andrew.