Current GCC seems to be worse in Spec95

[law@redhat.com]

  In message < 20010425082639.A7848@tornado.cygnus.com >you write:
  > On Wed, 25 Apr 2001, Andreas Jaeger wrote:
  > 
  > > 
  > > Looking at Diego's daily Spec tests at
  > > http://www.cygnus.com/~dnovillo/SPEC/sandbox-gcc/
  > > I noticed that the values for 126.GCC have been dropped during the
  > > last fortnight, it's now 2 points lower.
  > > 
  > > Any ideas why GCC is slower now?
  > > 
  > The slowdown started on 2001-04-18. These are the diffs for that
  > day:
  > 
  > ---------------------------------------------------------------------------
  > --
  > Tue Apr 17 21:41:11 2001  Jeffrey A Law  (law@cygnus.com)
  > 
  > 	* jump.c (mark_all_labels): Canonicalize the tail recursion
  > 	label attached to CALL_PLACEHOLDER insns.
  > 
  > 2001-04-17  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
  > 
  > 	* output.h (dump_flow_info, free_basic_block_vars,
  > 	which_alternative): Delete redundant prototypes.
  > 
  > 	* rtl.h (bss_section): Likewise.
  > 
  > 2001-04-17  Loren J. Rittle  <ljrittle@acm.org>
  > 
  > 	* dbxout.c (N_OPT): Define if not available from <stab.h>
  > 
  > 2001-04-17  John David Anglin  <dave@hiauly1.hia.nrc.ca>
  > 
  > 	* vax.h (PROMOTE_PROTOTYPES): Define for compatibility with system
  > 	libraries and native K&R compilers.
  > ---------------------------------------------------------------------------
  > --
  > 
  > The only one that seems relevant for SPEC is the first one. Note
  > that I have only quickly scanned the logs. I have not tested
  > anything yet.
I'd be flabbergasted if the jump.c change had any impact on performance, but
someone with more time should definitely do some investigation.  

jeff