OpenACC-Library-Interoperability

[Vikram Singh]

Well, I have already written code that I had originally wanted to do,
that is, use openacc with cublas.

And it works perfectly well with PGI compiler. I guess I will have to
check on how I would go about trying to run it using gfortran now that
host_data works using gfortran.

On Thu, May 12, 2016 at 6:34 PM, Thomas Schwinge
<thomas@codesourcery.com> wrote:
> Hi!
>
> On Fri, 15 Apr 2016 13:59:53 +0300, Vikram Singh <vikramsingh001@gmail.com> wrote:
>> I checked libgomp.oacc-c-c++-common/context-*.c, and they seem to be
>> exactly what I needed to start with.
>>
>> If I understand it correctly, I'll anyway be trying to implement
>> something similar with PGI compilers, so I'll have something as a
>> starting point.
>
> Any progress on that already?  I had a very quick look myself, but it's
> not as easy as I thought...  A "courageous" use of "use cublas" in
> Fortran code compiled with gfortran results in: "Fatal Error: Can't open
> module file 'cublas.mod' for reading".  The problem is: Fortran
> interfacing to C libraries (which cuBLAS is).
> <http://docs.nvidia.com/cuda/cublas/index.html#appendix-b-cublas-fortran-bindings>
> has some instructions how to do it.  Anyone got that to work already?
>
>> But again, it will need the OpenACC host_data construct to be setup
>> for gfortran to test.
>
> In
> <http://news.gmane.org/find-root.php?message_id=%3C2b4f59d5-be38-2814-27bb-73aa7ffb4b8f%40codesourcery.com%3E>,
> Chung-Lin has now posted a patch (pending review) that should make the
> OpenACC host_data construct usable in GCC Fortran.  (Problem discussed in
> <http://news.gmane.org/find-root.php?message_id=%3C878u0o6wwj.fsf%40kepler.schwinge.homeip.net%3E>
> before.)
>
> For reference:
>
>> On Fri, Apr 15, 2016 at 11:57 AM, Thomas Schwinge
>> <thomas@codesourcery.com> wrote:
>> > On Fri, 15 Apr 2016 11:35:06 +0300, Vikram Singh <vikramsingh001@gmail.com> wrote:
>> >> Yes, I came to the conclusion that host_data
>> >> would be the only way to do it in fortran.
>> >>
>> >> On the other hand, I though there were no plans to implement it in
>> >> gfortran 6 either
>> >
>> > I still hope we'll get this (that is, <https://gcc.gnu.org/PR70598>)
>> > fixed in time for the GCC 6.1 release.  I'll keep you posted.
>> >
>> >
>> >> > I'm copying Jim, who is the author of this chapter in the documentation
>> >> > as well as the libgomp/testsuite/libgomp.oacc-c-c++-common/context-*.c
>> >> > test cases, and much of the relevant libgomp code, too, and who should
>> >> > please correct me if I'm wrong.  I'll make a note for later, that we
>> >> > should translate the libgomp.oacc-c-c++-common/context-*.c test cases to
>> >> > Fortran, and also replicate them using the OpenACC host_data construct
>> >> > (like in libgomp/testsuite/libgomp.oacc-c-c++-common/host_data-1.c), and
>> >> > the same for the documentation you referenced.  (Vikram, please tell if
>> >> > you're interested to work on these items.)
>> >>
>> >> I am not sure I understand what you want me to work on exactly. I am
>> >> not really much of a C programmer, so I wouldn't be good at it. I
>> >> would still like to help.
>> >
>> > Sorry for being unclear.  My idea/question has been whether you're
>> > interested in helping by translating the documentation as well as the
>> > libgomp.oacc-c-c++-common/context-*.c test cases from C to Fortran (using
>> > the OpenACC host_data construct instead of the acc_* functions).  If yes,
>> > then that's great, if not, then one of us will do it at some point.
>
>
> Grüße
>  Thomas