Hi! On Mon, 16 Dec 2013 09:58:07 +0100, Jakub Jelinek <ja...@redhat.com> wrote: > On Fri, Dec 13, 2013 at 11:12:03AM +0100, Thomas Schwinge wrote: > > --- gcc/gimplify.c > > +++ gcc/gimplify.c > > @@ -129,6 +129,8 @@ struct gimplify_ctx > > bool in_cleanup_point_expr; > > }; > > > > +/* Variable remapping context. */ > > Please mention there OpenMP (and after OpenACC support is added, you can > tweak that to OpenMP/ACC or similar (in comments, not in diagnostics of > course).
> > -/* Create a new omp construct that deals with variable remapping. */ > > +/* Create a new variable remapping context for REGION_TYPE. */ > > Ditto, removing omp here loses important information, if you eventually > make it omp/oacc, it will be clear what you are talking about. This has been a tiny first attempt to follow up with what has been discussed at the GNU Tools Cauldron, that we intend to generalize the existing code, originally implemented for OpenMP, to be suitable not only for OpenMP but also OpenACC, Cilk Plus, and whichever other acceleration/parallelization scheme (general name yet to be found...). My rationale here (and sorry for not posting that: I again had falsely assumed this was obvious enough) was that while (I assume) this code had originally been written for OpenMP, and continues to be used for OpenMP, it is not specific to OpenMP, so to not induce that impression in the reader of the code, I elected a generic non-OpenMP name for it, »variable remapping context«. (But did not change the *omp* names of the datatypes/functions themselves, as discussed before: »names are sticky«.) Again, per the GNU Tools Cauldron discussion, my attempt was to do such refactoring/generalization incrementally, thus this first little patch. I can certainly add back OpenMP notes from where I removed them, but then I might as well just retract this patch, basically? Hmm. Grüße, Thomas
pgp08KJdiFGmj.pgp
Description: PGP signature