This is the mail archive of the
mailing list for the GCC project.
Re: [tree-ssa] RFC: Making control flow more explicit
- From: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- To: Dorit Naishlos <DORIT at il dot ibm dot com>
- Cc: Diego Novillo <dnovillo at redhat dot com>, gcc at gcc dot gnu dot org
- Date: Mon, 11 Aug 2003 21:45:28 +0200
- Subject: Re: [tree-ssa] RFC: Making control flow more explicit
- References: <OF78BC606E.2BF46595-ONC2256D7F.00162D5C-C2256D7F.0050EA64@telaviv.ibm.com>
> In the new lowered IR, would there be a simple high level way to detect
yes, see cfgloop.c.
> work with loops and loop nests, to duplicate loops, etc?
yes, in fact it will IMHO be much easier than now as you will be able
to just play with cfg in any way you want, without care of handling
the underlying structured expressions.
> Is there going to
> be something that marks that a certain CFG region is a loop / a loop with a
> single exit, etc? (I'm sure you had loop transformations in mind when you
> designed this low-GIMPLE IR, I'm trying to get an idea on how optimizations
> on loops would look like).
yes, again see cfgloop.c + cfgloopmanip.c. Also some stuff done on
rtlopt branch loop optimizer could be useful to check.