October 2008 - mpiwg-fortran - mailman3.cels.anl.gov

Re: [MPI3 Fortran] Variable length argument lists
by Jeff Squyres 21 Oct '08

21 Oct '08

Excellent; many thanks. On Oct 21, 2008, at 7:36 AM, Lionel, Steve wrote: > Fortran does not have variable-length argument lists. It does have > optional arguments. The only context in Fortran where one sees > anything > that looks like variable argument lists is in the MAX and MIN > intrinsics > which have an unspecified number of arguments designated A1, A2, A3, > etc. This cannot be expressed in Fortran source. As best as I can > tell, > the proper action is to remove the ",..." from the callback signature. > I would guess that this is intended to pass some sort of user > context to > the callback routine, which is a helpful idea. You could define an > optional CONTEXT argument of type C_PTR, but this would require > existing > applications to be rebuilt using a module providing an explicit > interface. Possibly more trouble than it's worth. > > It is also not correct to say "you can pass as many/few arguments as > you > want". While this may work with some implementations, with others it > does not. > > > Steve Lionel > Developer Products Division > Intel Corporation > Nashua, NH > > > > >> -----Original Message----- >> From: mpi3-fortran-bounces(a)lists.mpi-forum.org [mailto:mpi3-fortran- >> bounces(a)lists.mpi-forum.org] On Behalf Of Jeff Squyres >> Sent: Tuesday, October 21, 2008 7:23 AM >> To: MPI-3 Fortran working group >> Subject: [MPI3 Fortran] Variable length argument lists >> Importance: Low >> >> Fortran experts -- >> >> The MPI 2.1 spec lists some Fortran callback signatures as having >> variable arguments. I do believe that this is a holdover from a >> long, >> long time ago, and I believe that whoever wrote this section simply >> copied the C prototypes. If the spec is wrong, I'd like to get it >> fixed during the ongoing MPI-2.1 process. See MPI-2.1 p266 line 29, >> or http://www.mpi-forum.org/docs/mpi21-report-bw/node183.htm#Node183. >> >> The Fortran callback signature is expressed as: >> >> SUBROUTINE COMM_ERRHANDLER_FN(COMM, ERROR_CODE, ...) >> INTEGER COMM, ERROR_CODE >> >> 1. Does Fortran have variable length argument lists (akin to C >> varargs)? If so, was it introduced in a specific version of Fortran? >> (I realize that F77 compilers do no checking and you can pass as >> many/ >> few arguments as you want) >> >> 2. If Fortran has variable length argument lists, what is the correct >> syntax to express them? Is "..." correct? Or is there a different >> way? >> >> Thanks. >> >> -- >> Jeff Squyres >> Cisco Systems >> >> _______________________________________________ >> mpi3-fortran mailing list >> mpi3-fortran(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-fortran > > _______________________________________________ > mpi3-fortran mailing list > mpi3-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-fortran -- Jeff Squyres Cisco Systems

1 0

Re: [MPI3 Fortran] Variable length argument lists
by Lionel, Steve 21 Oct '08

21 Oct '08

Fortran does not have variable-length argument lists. It does have optional arguments. The only context in Fortran where one sees anything that looks like variable argument lists is in the MAX and MIN intrinsics which have an unspecified number of arguments designated A1, A2, A3, etc. This cannot be expressed in Fortran source. As best as I can tell, the proper action is to remove the ",..." from the callback signature. I would guess that this is intended to pass some sort of user context to the callback routine, which is a helpful idea. You could define an optional CONTEXT argument of type C_PTR, but this would require existing applications to be rebuilt using a module providing an explicit interface. Possibly more trouble than it's worth. It is also not correct to say "you can pass as many/few arguments as you want". While this may work with some implementations, with others it does not. Steve Lionel Developer Products Division Intel Corporation Nashua, NH > -----Original Message----- > From: mpi3-fortran-bounces(a)lists.mpi-forum.org [mailto:mpi3-fortran- > bounces(a)lists.mpi-forum.org] On Behalf Of Jeff Squyres > Sent: Tuesday, October 21, 2008 7:23 AM > To: MPI-3 Fortran working group > Subject: [MPI3 Fortran] Variable length argument lists > Importance: Low > > Fortran experts -- > > The MPI 2.1 spec lists some Fortran callback signatures as having > variable arguments. I do believe that this is a holdover from a long, > long time ago, and I believe that whoever wrote this section simply > copied the C prototypes. If the spec is wrong, I'd like to get it > fixed during the ongoing MPI-2.1 process. See MPI-2.1 p266 line 29, > or http://www.mpi-forum.org/docs/mpi21-report-bw/node183.htm#Node183. > > The Fortran callback signature is expressed as: > > SUBROUTINE COMM_ERRHANDLER_FN(COMM, ERROR_CODE, ...) > INTEGER COMM, ERROR_CODE > > 1. Does Fortran have variable length argument lists (akin to C > varargs)? If so, was it introduced in a specific version of Fortran? > (I realize that F77 compilers do no checking and you can pass as many/ > few arguments as you want) > > 2. If Fortran has variable length argument lists, what is the correct > syntax to express them? Is "..." correct? Or is there a different > way? > > Thanks. > > -- > Jeff Squyres > Cisco Systems > > _______________________________________________ > mpi3-fortran mailing list > mpi3-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-fortran

1 0

[MPI3 Fortran] Variable length argument lists
by Jeff Squyres 21 Oct '08

21 Oct '08

Fortran experts -- The MPI 2.1 spec lists some Fortran callback signatures as having variable arguments. I do believe that this is a holdover from a long, long time ago, and I believe that whoever wrote this section simply copied the C prototypes. If the spec is wrong, I'd like to get it fixed during the ongoing MPI-2.1 process. See MPI-2.1 p266 line 29, or http://www.mpi-forum.org/docs/mpi21-report-bw/node183.htm#Node183. The Fortran callback signature is expressed as: SUBROUTINE COMM_ERRHANDLER_FN(COMM, ERROR_CODE, ...) INTEGER COMM, ERROR_CODE 1. Does Fortran have variable length argument lists (akin to C varargs)? If so, was it introduced in a specific version of Fortran? (I realize that F77 compilers do no checking and you can pass as many/ few arguments as you want) 2. If Fortran has variable length argument lists, what is the correct syntax to express them? Is "..." correct? Or is there a different way? Thanks. -- Jeff Squyres Cisco Systems

1 0

Re: [MPI3 Fortran] No Fortran next week
by Craig Rasmussen 17 Oct '08

17 Oct '08

The world Fortran standards body is meeting in Japan next month. At that time, there will be several votes taken on what will go into the next standard. Until we get feedback from the Fortran committee, we can't do much more on the Fortran bindings. The December meeting will be a good time to discuss where the Fortran standard is going and start to finalize the Fortran MPI bindings. I'm hoping to be able to attend this meeting. Cheers, Craig On Oct 17, 2008, at 8:28 AM, Jeff Squyres wrote: > I don't know if Craig got around to mailing this or not, but I just > wanted to let everyone know that Craig won't be there next week, and > I don't think we have anything new to discuss. So there won't be a > Fortran sub-group next week. > > > I think that this is a very high-level list of the open issues: > > - fixing/working around/making less sucky/whatever buffer aliasing > issues for non-blocking MPI communication > - using more advanced Fortran language features > > Right? > > -- > Jeff Squyres > Cisco Systems > > _______________________________________________ > mpi3-fortran mailing list > mpi3-fortran(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-fortran

1 0

[MPI3 Fortran] No Fortran next week
by Jeff Squyres 17 Oct '08

17 Oct '08

I don't know if Craig got around to mailing this or not, but I just wanted to let everyone know that Craig won't be there next week, and I don't think we have anything new to discuss. So there won't be a Fortran sub-group next week. I think that this is a very high-level list of the open issues: - fixing/working around/making less sucky/whatever buffer aliasing issues for non-blocking MPI communication - using more advanced Fortran language features Right? -- Jeff Squyres Cisco Systems

1 0