This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Simple recompute_reg_usage cleanup
- To: Joern Rennecke <amylaar at cygnus dot co dot uk>
- Subject: Re: Simple recompute_reg_usage cleanup
- From: Michael Hayes <m dot hayes at elec dot canterbury dot ac dot nz>
- Date: Sat, 4 Dec 1999 16:25:13 +1300 (NZDT)
- Cc: m dot hayes at elec dot canterbury dot ac dot nz (Michael Hayes), hubicka at atrey dot karlin dot mff dot cuni dot cz, gcc-patches at gcc dot gnu dot org, rth at cygnus dot com
- References: <14408.29602.60982.562338@ongaonga.elec.canterbury.ac.nz><199912040307.DAA12304@phal.cygnus.co.uk>
Joern Rennecke writes:
> > Yes, but we could also search back before the loop header to see if
> > the exit test has been duplicated.
>
> Which means you have to check the tests for equivalence. This sounds
> like a lot of work to get information that we get for free with the
> current VTOP scheme.
Yes, I agree that this is not easy; especially after strength
reduction changes the exit test. It may be better for jump.c to add a
note to the exit test to say that it has been duplicated.
The problem I see with the current scheme is that we might get
confused as to which natural loop the VTOP note refers to. This won't
be a problem if jump.c uses the CFG when it decides to duplicate the
exit test. In the interim, I'm looking at ignoring natural loops
that do not match the LOOP_BEG/LOOP_END notes.
Michael.