Hi Jakub! On Fri, 23 Oct 2015 15:58:47 +0200, Jakub Jelinek <ja...@redhat.com> wrote: > On Fri, Oct 23, 2015 at 03:38:31PM +0200, Thomas Schwinge wrote: > > On Fri, 23 Oct 2015 12:48:23 +0200, I wrote: > > > On Fri, 23 Oct 2015 10:07:49 +0200, I wrote: > > > > On Mon, 19 Oct 2015 17:54:46 +0200, I wrote: > > > > > [...] Jakub's big "Merge from > > > > > gomp-4_1-branch to trunk", > > > > > <http://news.gmane.org/find-root.php?message_id=%3C20151013191214.GL478%40tucnak.redhat.com%3E>. > > > > > Instead of attempting to merge that one in one go -- that is, to avoid > > > > > having to deal with a ton of merge conflicts at once, and to allow for > > > > > easier understanding of individual changes/regressions -- in the > > > > > following I'll gradually merge individual "blocks" of all the > > > > > gomp-4_1-branch changes into gomp-4_0-branch. > > > > Amongst other things, the last batch includes support for struct elements > > in OpenMP "map", "to", "from" clauses. Again, I have not yet assessed to > > which extent this will be useful to support related (future) OpenACC > > functionality. > > I bet that the OpenACC standard would need to say something about those > before you can start using them.
:-) That's why I here said *future* OpenACC functionality. > That said, I think more interesting to OpenACC might be the > firstprivatization of scalars on the target region (so that instead of > mapping the arguments that are newly mapped and only copied from the host to > device, you can use a cheaper firstprivatization). If OpenACC requires all > user vars to be mapped, you could still consider that e.g. for artificial > variables created by the compiler. Right, using that "firstprivate" functionality is one of the things that I noted down while working through your changes; should be a nice optimization. Grüße Thomas
signature.asc
Description: PGP signature
