[Bug tree-optimization/50693] Loop optimization restricted by GOTOs

["jakub at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50693

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |jakub at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #24 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-11-04 17:18:34 UTC ---
Created attachment 25721
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25721
gcc47-pr50693.patch

Here is that idea implemented.  On the gimple/expansion side it was trivial,
the only thing that complicated it is to avoid -fcompare-debug differences.
NOTE_INSN_DELETED_LABELs use the same numbers as other labels, unique for the
whole CU, so if we start to allocate from that number pool for these labels
that will be only present with -g (since they live in debug stmts and later on
in DEBUG_INSNs), we'll get label number differences, and furthermore for darwin
does some terrible hacks to workaround the mess called MachO it could even
result in different code genration.
So I decided to add NOTE_INSN_DELETED_LABEL variant which will use a different
label namespace.