This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [hsa merge 09/10] Majority of the HSA back-end


On Fri, Jan 15, 2016 at 04:08:14PM +0100, Martin Jambor wrote:
> We don't error, apart from issuing a warning we basically ignore them.
> I believe we can do it even in the long term and that it is in fact
> useful because the standard says that the "effect" if these routines
> is "unspecified" if they get called from a target region.
> 
> Perhaps this is even something we should warn about earlier in omp
> lowering/expansion.

Well, only some of the omp.h functions are not allowed to be called from 
the target regions, others are.
For the others that have unspecified behavior, there is always the question
when it is desirable to warn.  In target construct body it might be a
warning candidate, the only possibility that it is not invoking unspec
behavior is if the target construct is not encountered, if it is in dead
code in that body, or just never encountered.
But if you have declare target routine, it is more controversial to warn,
because the routine can be run both on host (where it is fine) and on target
(where it is not), whether it calls the argument e.g. could depend on some
parameter or result of some function (say check whether it is in offloaded
region).

Anyway, thanks for fixing this, the patch is ok for trunk.  And after the
commit you're the maintainer, so it is up to you to review further changes
to it.  Please keep it nicely and consistently formatted in the future ;)

	Jakub


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]