Jan,
You can use MPI_Type_match_size (
https://www.open-mpi.org/doc/v2.0/man3/MPI_Type_match_size.3.php) to find
the MPI type that meets certain requirements.
George.
On Wed, Dec 21, 2016 at 3:16 AM, 🐋 Jan Hegewald
wrote:
> Hi Jeff et al,
>
> > On 20 Dec 2016, at 20:23, Jeff Squyres (jsquy
Hi Jeff et al,
> On 20 Dec 2016, at 20:23, Jeff Squyres (jsquyres) wrote:
>
> Fair enough.
>
> But be aware that if you are building MPICH with double-width Fortran types
> and single-width C types, you might want to verify that MPICH is actually
> working properly.
yes. This episode certain
Fair enough.
But be aware that if you are building MPICH with double-width Fortran types and
single-width C types, you might want to verify that MPICH is actually working
properly.
Open MPI is refusing to configure because for each Fortran type X, it looks for
a corresponding C type Y. If it
Hi Gilles,
> On 20 Dec 2016, at 05:31, Gilles Gouaillardet wrote:
>
> btw, do you really want 32 bytes complex in Fortran ?
that is not for me to decide. I also do not want to change the default build
just because of the open-mpi flags.
>
> if yes, then you should use an other compiler (suc
Jan,
so it seems (the version of) your clang compiler does not support 32
bytes complex,
and hence i am not surprised Open MPI cannot even be configure'd.
btw, do you really want 32 bytes complex in Fortran ?
if yes, then you should use an other compiler (such as gcc/g++)
if not, then you
> On 19 Dec 2016, at 15:19, Gilles Gouaillardet
> wrote:
>
> Jeff,
>
> I am not aware of such flag for C/C++ compilers.
>
> Jan,
>
> I noticed
> checking size of Fortran DOUBLE PRECISION... 16
> At first glance, this looks surprising to me
> Which compiler (vendor and version) are you using
Hi Jeff and others,
> On 19 Dec 2016, at 14:58, Jeff Squyres (jsquyres) wrote:
>
> You'll need to set CFLAGS and CXXCFLAGS to the equivalent, too.
>
> For example:
>
> ./configure FCFLAGS=-fdefault-real-8 CFLAGS=-fdefault-real-8
> CXXCFLAGS=-fdefault-real-8 ...
>
> (I don't know if -fdefault
Jeff,
I am not aware of such flag for C/C++ compilers.
Jan,
I noticed
checking size of Fortran DOUBLE PRECISION... 16
At first glance, this looks surprising to me
Which compiler (vendor and version) are you using ?
Cheers,
Gilles
On Monday, December 19, 2016, Jeff Squyres (jsquyres)
wrote:
You'll need to set CFLAGS and CXXCFLAGS to the equivalent, too.
For example:
./configure FCFLAGS=-fdefault-real-8 CFLAGS=-fdefault-real-8
CXXCFLAGS=-fdefault-real-8 ...
(I don't know if -fdefault-real-8 is the correct flags for your C and C++
compilers -- you might need to poke through your do
> On 16 Dec 2016, at 20:42, Jeff Squyres (jsquyres) wrote:
>
> If you don't build Open MPI and your application with the same compiler
> flags, weird things like this are bound to happen. Make sure Open MPI is
> built the same way.
I have no success building open-mpi with -fdefault-real-8:
> On 16 Dec 2016, at 20:42, Jeff Squyres (jsquyres) wrote:
>
> If you don't build Open MPI and your application with the same compiler
> flags, weird things like this are bound to happen. Make sure Open MPI is
> built the same way.
Yes, I just did the default install with homebrew. I also ju
If you don't build Open MPI and your application with the same compiler flags,
weird things like this are bound to happen. Make sure Open MPI is built the
same way.
Also, note that the formal signature for MPI_Wtime is to return a DOUBLE
PRECISION. REAL(KIND=8) is common, but it is not techni
I found the problem. It is in the build process which passes -fdefault-real-8
(among other flags) to the compiler:
mpif90 -fdefault-real-8 foo.F90
➜ mpirun -n 1 ./a.out
0.0.
Cheers,
Jan
> On 16 Dec 2016, at 19:16, Jeff Squyres (jsquyres) wrote:
>
> On Dec 16, 2016, at 12:48 PM, 🐋 Jan Hegewald wrote:
>>
>> Your example works fine here too.
>> I call Wtime from Fortran code. But if I do the same as in your C example
>> 1), it seems to work as well.
>> I am puzzled as to what i
On Dec 16, 2016, at 12:48 PM, 🐋 Jan Hegewald wrote:
>
> Your example works fine here too.
> I call Wtime from Fortran code. But if I do the same as in your C example 1),
> it seems to work as well.
> I am puzzled as to what is causing this. On several HPC clusters everything
> is fine. Only on
> On 16 Dec 2016, at 18:02, Jeff Squyres (jsquyres) wrote:
>
> That's odd; that doesn't happen to me.
>
> E.g., with this program running on MacOS 10.12.1 with Open MPI 2.0.1, I see a
> different value for MPI_Wtime() between successive calls:
>
> -
> $ cat foo.c
Your example works fine
That's odd; that doesn't happen to me.
E.g., with this program running on MacOS 10.12.1 with Open MPI 2.0.1, I see a
different value for MPI_Wtime() between successive calls:
-
$ cat foo.c
#include
#include
#include
int main(int argc, char *argv[])
{
MPI_Init(NULL, NULL);
double
17 matches
Mail list logo
