Sorry - I do not see how that solves the problem. Compiled application code allocates some number of bytes based on a structure declaration in C or a constant (MPI_STATUS_SIZE) in Fortran. Each byte in the structure is required because the MPI library has a use for it. There will not be spare bytes that can be used to cache the extra baggage. Dick Treumann - MPI Team IBM Systems & Technology Group Dept X2ZA / MS P963 -- 2455 South Road -- Poughkeepsie, NY 12601 Tele (845) 433-7846 Fax (845) 433-8363 From: Greg Bronevetsky <[email protected]> To: "MPI 3.0 Fault Tolerance and Dynamic Process Control working Group" <[email protected]>, "MPI 3.0 Fault Tolerance and Dynamic Process Control working Group" <[email protected]> Date: 03/24/2010 12:36 PM Subject: Re: [Mpi3-ft] FT Working Group Teleconf (today) Sent by: [email protected]

I agree with Dick. We're getting very strong ISV feedback that an MPI-3 compatibility break will constitute a serious adoption barrier for them.

The other option would be to have an extra function like MPI_Get_count that extracts internal information from MPI_Status objects. Can that be implemented without breaking compatibility? Greg Bronevetsky Lawrence Livermore National Lab (925) 424-5756 [email protected] http://greg.bronevetsky.com _______________________________________________ mpi3-ft mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft