[mpich-discuss] How does the fortran binding work?

Ruini Xue xueruini at gmail.com
Wed Jul 30 20:19:53 CDT 2008


I failed to compile MPICH2 from its source in VS 2005. Could the develop
team provide a debug version? Thus we can use a simple fortran app to
inspect the real case: set a breakpoint on MPI_BCAST() and print the
callstack.

Best

Andrew

On Thu, Jul 31, 2008 at 9:17 AM, Ruini Xue <xueruini at gmail.com> wrote:

> Here is a link which illustrate this topic:
> http://www.netlib.org/utk/papers/mpi-book/node190.html
>
> All C version MPI_XXXX() are defined as *weak symbols* to PMPI_XXXX(),
> which means that if there is no MPI_XXX() definition, the
> user application would use the coresponding PMPI_XXX(). The figure also
> shows that in normal case (without profiling), PMPI_Bcast()
> do all the real work (see MPI_Bcast()--->PMPI_Bcast()) because MPI_Bcast()
> is not defined yet, otherwise (with profiling), since MPI_Send() is provided
> by the profiling lib, use app would be linked to the MPI_Send() in profiling
> lib, which in turn calls the original PMPI_Send(). So, I think it's
> PMPI_XXX() do all the work internally, both C and Fortran bindings calls
> PMPI_XXX().
>
> Now let check code. There a piece of code from: src/mpi/pt2pt/bsend.c
> ///////////////////////////////////////////////////////////
> /* -- Begin Profiling Symbol Block for routine MPI_Bsend */
> #if defined(HAVE_PRAGMA_WEAK)
> #pragma weak MPI_Bsend = PMPI_Bsend
> #elif defined(HAVE_PRAGMA_HP_SEC_DEF)
> #pragma _HP_SECONDARY_DEF PMPI_Bsend  MPI_Bsend
> #elif defined(HAVE_PRAGMA_CRI_DUP)
> #pragma _CRI duplicate MPI_Bsend as PMPI_Bsend
> #endif
> /* -- End Profiling Symbol Block */
>
> /* Define MPICH_MPI_FROM_PMPI if weak symbols are not supported to build
>    the MPI routines */
> #ifndef MPICH_MPI_FROM_PMPI
> #undef MPI_Bsend
> #define MPI_Bsend PMPI_Bsend
>
> #endif
>
> #undef FUNCNAME
> #define FUNCNAME MPI_Bsend
> //////////////////////////////////////////////////////////////////////////
>
> It seems that if #pragma weak is supported, MPI_Bsend is weakly defined to
> PMPI_Bsend, otherwise MPI_Bsend is defined to
> PMPI_Bsend via macro. That's to say, MPI_Bsend always does nothing, but is
> mapped to PMPI_Bsend. This is the C binding version.
>
> For fortran binding, check src/binding/f77/bsendf.c:
> //////////////////////////////////////////////////////////////////////
> FORT_DLL_SPEC void FORT_CALL mpi_bsend_ ( void*v1, MPI_Fint *v2, MPI_Fint
> *v3, MPI_Fint *v4, MPI_Fint *v5, MPI_Fint *v6, MPI_Fint *ierr ){
>     *ierr = MPI_Bsend( v1, *v2, (MPI_Datatype)(*v3), *v4, *v5,
> (MPI_Comm)(*v6) );
> }
> /////////////////////////////////////////////////////////////////////
>
> It seems that the fortran binding calls the C binding directly, but
> actually MPI_Bsend() is PMPI_Bsend().
>
> I am not sure whether I describe and analysis it accurately.
>
>
> Best
>
> Andrew
>
> On Thu, Jul 31, 2008 at 2:11 AM, Jayesh Krishna <jayesh at mcs.anl.gov>wrote:
>
>>  >> Looks like the Fortran interface does call the MPI version of the C
>> function.
>>
>>   Only if the call is from PMPI_* funcs. (If the MPI call is not from
>> PMPI, MPI_* funcs get defined to PMPI_*)
>>
>> >> If the fortran binding calls PMPI version of C function...
>>
>>   In windows this currently does not work (I guess, it is due to the
>> different calling convention/name mangling by the mpe lib and the fortran
>> lib in windows) :( Defenitely a TODO item :)
>>
>> Regards,
>> Jayesh
>>
>> -----Original Message-----
>> From: owner-mpich-discuss at mcs.anl.gov [
>> mailto:owner-mpich-discuss at mcs.anl.gov <owner-mpich-discuss at mcs.anl.gov>]
>> On Behalf Of Rajeev Thakur
>> Sent: Wednesday, July 30, 2008 12:42 PM
>> To: mpich-discuss at mcs.anl.gov
>> Subject: RE: [mpich-discuss] How does the fortran binding work?
>>
>> Looks like the Fortran interface does call the MPI version of the C
>> function.
>>
>>
>> > -----Original Message-----
>> > From: owner-mpich-discuss at mcs.anl.gov
>> > [mailto:owner-mpich-discuss at mcs.anl.gov<owner-mpich-discuss at mcs.anl.gov>]
>> On Behalf Of Anthony Chan
>> > Sent: Wednesday, July 30, 2008 12:01 PM
>> > To: mpich-discuss at mcs.anl.gov
>> > Subject: Re: [mpich-discuss] How does the fortran binding work?
>> >
>> > If the fortran binding calls PMPI version of C function, can the
>> > fortran MPI program still be profiled ?
>> >
>> > ----- "Rajeev Thakur" <thakur at mcs.anl.gov> wrote:
>> >
>> > > The Fortran binding calls the PMPI version of the C function.
>> > >
>> > > Rajeev
>> > >
>> > > > -----Original Message-----
>> > > > From: owner-mpich-discuss at mcs.anl.gov
>> > > > [mailto:owner-mpich-discuss at mcs.anl.gov<owner-mpich-discuss at mcs.anl.gov>]
>> On Behalf Of Anthony Chan
>> > > > Sent: Wednesday, July 30, 2008 9:46 AM
>> > > > To: mpich-discuss at mcs.anl.gov
>> > > > Cc: mpich-discuss at mcs.anl.gov
>> > > > Subject: Re: [mpich-discuss] How does the fortran binding work?
>> > > >
>> > > >
>> > > > No.  PMPI is the MPI profiling interface and is standardized by
>> > > > the MPI forum.
>> > > > Fortran/C++ MPI binding just calls the C MPI routine.
>> > > >
>> > > > A.Chan
>> > > >
>> > > > ----- "Ruini Xue" <xueruini at gmail.com> wrote:
>> > > >
>> > > > > On Wed, Jul 30, 2008 at 8:54 PM, Dave Goodell <
>> > > > goodell at mcs.anl.gov >
>> > > > > wrote:
>> > > > >
>> > > > >
>> > > > > Hi Andrew,
>> > > > >
>> > > > > Yes, the fortran and c++ bindings are just wrappers over the C
>> > > > > bindings. The C binding implements the functionality directly.
>> > > > >
>> > > > > Are they wrappers of MPI_XXX() or PMPI_XXX()?
>> > > > >
>> > > > > Best
>> > > > >
>> > > > > Andrew
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > > -Dave
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > > On Jul 30, 2008, at 6:11 AM, Ruini Xue wrote:
>> > > > >
>> > > > >
>> > > > >
>> > > > > Hello, everyone,
>> > > > >
>> > > > > I am working with MPICH2 on Windows platfrom. The fortran MIP
>> > > > > application is linked to fmpich.lib, and fmpich2.dll is
>> > loaded in
>> > > > > runtime, which in turn loads mipch2mpi.dll.
>> > > > > mpich2mpi.dll contains all MPI APIs which exported C
>> > > > interface. What I
>> > > > > want to know is: what is the 'underlying' relation between
>> > > > fmpich2.dll
>> > > > > and mpich2mip.dll? Do all
>> > > > > Fortran MPI calls are converted to c bindings in mpich2mpi.dll?
>> > > For
>> > > > > example, does MPI_BCAST() in fmich2.dll directly calls
>> > > > MPI_Bcast() in
>> > > > > mpich2mpi.dll? Is MPI_BCAST() just a stub while MPI_Bcast() do
>> > > the
>> > > > > real job?
>> > > > >
>> > > > > Thanks
>> > > > >
>> > > > > Andrew
>> > > >
>> > > >
>> >
>> >
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20080731/9e17a330/attachment.htm>


More information about the mpich-discuss mailing list