This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Add -fsection-anchors (4.2 project)
Richard Sandiford wrote:
> Well, my last reply was largely trying to argue that the optimisation
> _can_ hurt you, even with base+offset addressing. The point is that not
> every address calculation is used directly for a load and store. It might
> be used to compute another address, such as one involving a variable index.
> Or it might be hoisted, stored, or passed to a function.
OK, now I understand what you meant.
> I'm not sure that really helps, but does what I wrote in my last
> reply make any more sense after the above?
Yes, thanks.
> I don't really understand what you're asking for here. Grouping objects
> into blocks -- in a haphazard order -- doesn't have any direct advantage.
> It only helps if you make use of the information somehow.
Indeed.
> If some future pass
> wants to make use of the information, it's simply a case of adding the
> appropriate condition to use_object_blocks_p.
Right, that's good.
--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713