From: snir@XXXXXXXXXX X-Lotus-FromDomain: IBMUS To: Nicholas Nevin - Sun HPC High Performance Computing cc: mpi-core@XXXXXXXXXXX Date: Sun, 28 Mar 1999 19:24:47 -0500 Subject: Re: clarification on inter-language interoperability of attributes Mime-Version: 1.0 Content-type: text/plain; charset=us-ascii Content-Disposition: inline Sender: owner-mpi-core@XXXXXXXXXXX Precedence: bulk X-UIDL: c5634f68e673765c6a8e57ac994a4a1e Sorry -- I was wrong. Internally, Fortran attributes are of type MPI_ADDRESS_KIND, indeed. If such an attribute is retrieved in C, then MPI_xxx_get_attr will return a pointer (of type void*) that will point to an integer of suitable size. Where I said INTEGER, below, I should have said INTEGER(KIND=MPI_ADDRESS_KIND). Marc Snir Senior Manager, Scalable Parallel Systems IBM T J Watson Research Center, PO Box 218, Yorktown Heights 10598 Tel: 914 945 3204 (862 3204) Fax: 914 945 4425 (862 44245) URL: http://www.research.ibm.com/people/s/snir Nicholas Nevin - Sun HPC High Performance Computing on03/26/99 04:38:51 PM Please respond to Nicholas Nevin - Sun HPC High Performance Computing To: Marc Snir/Watson/IBM@IBMUS cc: mpi-core@XXXXXXXXXXX Subject: Re: clarification on inter-language interoperability of attributes > From: snir@XXXXXXXXXX > X-Lotus-FromDomain: IBMUS > To: Nicholas Nevin - Sun HPC High Performance Computing > cc: gropp@XXXXXXXXXXXX mpi-co > Date: Fri, 26 Mar 1999 16:20:21 -0500 > Subject: Re: clarification on inter-language interoperability of attributes > Content-Disposition: inline > > > MPI_xxx_get_attr returns a void* value in "attr". If the attribute you accessed > is a Fortran attribute then (int*)attr will point to an INTEGER. > MPI_ADDRESS_KIND is irrelevant. > > > I'm afraid I have to disagree. MPI_ADDRESS_KIND is relevant. Here is the binding for MPI_COMM_SET_ATTR. MPI_COMM_SET_ATTR(COMM, COMM_KEYVAL, ATTRIBUTE_VAL, IERROR) INTEGER COMM, COMM_KEYVAL, IERROR INTEGER(KIND=MPI_ADDRESS_KIND) ATTRIBUTE_VAL ^^^^^^^^^^^^^^^^ ????? -nick