MPI and UNICODE

• Question

• 

  

  0

  Sign in to vote

  Hi all.  I'm curious what the workarounds are for compiling UNICODE apps with MSMPI.  In short, here's the problem.  A typical windows app is defined like this:

   

  int _tmain(int argc, _TCHAR* argv[ ])

  {

   .

   .

   .

  }

   

  If UNICODE is defined, then _TCHAR becomes a WCHAR, i.e. a UNICODE character.  In this case, the call to MPI_Init fails to compile, since MPI is defined assuming ANSI characters, not UNICODE:

   

    MPI_Init(&argc, &argv);   // type error, since prototype is int MPIAPI MPI_Init(int *argc, char ***argv)

   

  Is this a known limitation?  Or maybe I'm missing something?  I have a workaround, which amounts to overriding MPI_Init and replacing it with a UNICODE-compatible version. 

   

  Thanks for any ideas/pointers.  Cheers,

   

    - joe

  Tuesday, March 4, 2008 1:58 AM