Re: Save memory with ppc elf shared libraries.

[Jeffrey A Law <law at cygnus dot com>]

  In message <199812091112.WAA25840@geoffk.wattle.id.au>you write:
  > Now, in this case, most symbolic addresses are already in memory.  But
  > addresses of the form
  > 
  > (const (plus (sym_ref "foo") 10))
  > 
  > are not put in the GOT.
  > 
  > Can reload ever try to put this in memory?  Even if it can load
  > (sym_ref "foo") from memory and perform the addition in the same
  > register it performed the load from?
Yes.  See the calls to force_const_mem in reload.c.  Been there, done that.


  > [Of course, reload can't ever try the same thing if the second operand
  > isn't a const_int, because the linker can't handle it.]
Right.

  > LEGITIMIZE_ADDRESS seems to always do such computations in registers.
But we're talking about a completely different part of the compiler.
LEGITIMIZE_ADDRESS is a totally target dependent hunk of code.  The code
which forces constants into memory is generic code in reload.


  > Perhaps, though, it might be best to change rs6000_select_rtx_section
  > to be on the safe side.  Something like:
  > 
  > if (flag_pic == 1 &&
  >     (GET_CODE (x) == SYMBOL_REF
  >      || (GET_CODE (x) == CONST && GET_CODE (XEXP (x, 0)) == PLUS
  >          && GET_CODE (XEXP (XEXP (x, 0), 0)) == SYMBOL_REF)
  >      || GET_CODE (x) == LABEL_REF))
  >   data_section ();
  > else if (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P (x))
  >   toc_section ();
  > else
  >   const_section ();
  > 
  > But I can't test this, because I have no examples of a case when it
  > might ever be used...
They can be difficult to trigger since they involve reload.

  > There really should be a macro to mean "includes symbol reference",
  > that is the condition above.
Yes.  Though it also needs to check for LABEL_REFs in precisely the same
way you check for SYMBOL_REFs.  I would support putting the code into a
generic place in the compiler since more than just the ppc can use it.

My recommendation is to have a function or macro which returns nonzero if
the given rtx contains a SYMBOL_REF or LABEL_REF.  Then change
SELECT_RTX_SECTION to use the data section if flag_pic is nonzero (flag_pic
can have 3 values 0, 1 and 2.  1 is used for -fpic   2 is used for -fPIC.

Combine that with your previous patch and I think we've got the right batch
of changes.

jeff