This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: new timevars


> Jan Hubicka <jh at suse dot cz> writes:
> 
> > Sat Feb 22 12:53:13 CET 2003  Jan Hubicka  <jh at suse dot cz>
> > 	* Makefile.in (jump.o, regclass.o, alias.o): Add dependency on timevar.h
> > 	* alias.c: Include timevar.h
> > 	(init_alias_analysis):  Set timevar
> > 	* jump.c: Include timevar.h
> > 	(rebuild_jump_labels):  Set timevar
> > 	* regcalss.c: Include timevar.h
> > 	(reg_scan):  Set timevar
> > 	* timevar.def (TV_ALIAS_ANALYSIS, TV_REG_SCAN, TV_REBUILD_JUMP): New
> 
> Hmm.  Is init_alias_analysis really the only place that should be
> put under TV_ALIAS_ANALYSIS?  And why is TV_REBUILD_JUMP distinct from
Dificult question.  We do have the functions like canon_rtx that do use
the alias analysis, but I am not sure we want to mark them.  I was
mainly interested in how much time we consume in collecting the info, as
we do so many times in the copilation (and the amount can clearly be
reduced)
> TV_JUMP?
Because once jump pass is dead we don't have anything real fro TV_JUMP.
TV_JUMP now includes many things, most of them not jump related, so I
don't want to use it anymore.

Honza
> 
> zw


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]