Perhaps if we generalized the part that adds metadata to the loop so attributes and non-attributes (openmp stmts) can use the same mechanism? I’m not sure how other aspects are compatible. I would have to think about it.
I’ll work on a patch to merge these mechanisms in a follow-up commit. Tyler On May 22, 2014, at 2:27 AM, Alexander Musman <[email protected]> wrote: > I've just committed my change with emission of the loop metadata for > vectorizer for 'omp simd' (in r209411) and I think that in some follow-up > patch it would be good to do a change to use the same code path in both (i.e. > use CGLoopInfo to emit loop metadata for all vectorizing pragmas). Actually, > if '#pragma loop vectorize' will get a clause for marking loads/stores > loop-parallel, it is good idea to reuse it anyway. > > Alexander > > > 2014-05-20 20:12 GMT+04:00 Tyler Nowicki <[email protected]>: > Ping. > > On May 14, 2014, at 10:05 AM, Tyler Nowicki <[email protected]> wrote: > > > Hi Alexey, Aaron, > > > > Thanks for the review. Richard and John please review the attached patch. > > > > Tyler > > > > <pragma_loop-svn.patch> > > > > On May 13, 2014, at 9:51 PM, Alexey Bataev <[email protected]> wrote: > > > >> I think CodeGen part of the patch should be reviewed by John McCall. > > _______________________________________________ > cfe-commits mailing list > [email protected] > http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits >
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
