[mpich-discuss] Cannot build mpich2-1.0.8p1 (nemesis) with PGI8.0-4 on Linux x86_64

[Pavan Balaji]

Right now this is restricted to compilers producing compatible object 
files with respect to a few things, listed here: 
http://wiki.mcs.anl.gov/mpich2/index.php/Frequently_Asked_Questions#Building_MPICH2

But a perfect solution for all compilers is hard, as you pointed out. 
But having something is better than nothing :-). We are putting together 
a list of tested compilers which we know work with each other. If you 
are outside that list, then there are no guarantees, obviously.

  -- Pavan

Jeff Squyres (jsquyres) wrote:
> F90 does weird name mangling as well.  But I'm not familiar with mpich's 
> f90 support - are there no overloaded functions?  Or have you tested 
> multiple f90 compilers to know that they mangle the same way?  (I freely 
> admit to not having tested f90 compilers to know)
> 
> Does mpich fix the ".true. value" issue at run time somehow?  That 
> particular problem has vexed me for a while...
> 
> Also, I never figured out how to make the C constants 
> mpi_f_status(es)_ignore able to compare properly in fortran apps that 
> were compiled with a different name mangling scheme - did you figure out 
> a solution for that as well?
> 
> -jms
> Sent from my PDA.  No type good.
> 
> ----- Original Message -----
> From: mpich-discuss-bounces at mcs.anl.gov <mpich-discuss-bounces at mcs.anl.gov>
> To: mpich-discuss at mcs.anl.gov <mpich-discuss at mcs.anl.gov>
> Sent: Wed Apr 01 19:30:38 2009
> Subject: Re: [mpich-discuss] Cannot build mpich2-1.0.8p1 (nemesis) with 
> PGI8.0-4 on Linux x86_64
> 
> 
>  > Do you mean Fortran 77 only, or will this include F90 (mpif90)?
> 
> It should work for both. C++ is the only problematic area because of its
> weird name mangling.
> 
>   -- Pavan
> 
> --
> Pavan Balaji
> http://www.mcs.anl.gov/~balaji
> 

-- 
Pavan Balaji
http://www.mcs.anl.gov/~balaji