February 2008 - Mpich2-dev - mailman3.cels.anl.gov

Re: [mpich2-dev] Fw: 32-bit ROMIO: What is the best solution?
by William Gropp 29 Feb '08

29 Feb '08

Any place where sizeof(int*) == sizeof(int) is assumed is just broken - if its easy to separate those out, we'll just apply a fix for those, since that's needed for any platform and under all conditions. The sizeof(MPI_Aint) == sizeof(void*) assumption will require more care. Bill On Feb 29, 2008, at 11:08 AM, Brian Smith wrote: > > That's basically what we are doing. The problem is, there are a > large number of places throughout the code (not just in ROMIO) > where there are casts that assume MPI_Aint == sizeof(void*)==sizeof > (int*)==sizeof(int), so we are walking through all that code and > seeing what the casts are doing and fixing them as appropriate. William Gropp Paul and Cynthia Saylor Professor of Computer Science University of Illinois Urbana-Champaign

3 2

Fw: 32-bit ROMIO: What is the best solution?
by Bob Cernohous 29 Feb '08

29 Feb '08

I'm moving a discussion here from the IBM/Argonne open source DCMF mailing list (http://dcmf.anl-external.org/wiki) Very briefly, we're making MPI_Aint into a 64 bit signed value on 32-bit MPICH2/ROMIO platforms. This will fix problems with files larger than 2G (and possibly other 4G virtual address problems). We'd like to start sending fix patches to the mpich-dev community for comment. The code base that we're patching is changing under us. So our patches might not apply properly for members of this list. When we're done, we will have a patch for the current release of MPICH2. For now, we'd like to provide these patches to the community for information and discussion. If you're interested, there might be more technical history in the DCMF archives. Otherwise, start the discussion here. bobc(a)us.ibm.com wrote on 02/27/2008 04:19:29 PM: > > bobc(a)us.ibm.com wrote on 02/08/2008 02:41:40 PM: > > >We all realize there are issues with MPI_Aint's in 32 bit implementations. > >Using signed 32 bit addresses along with 64 bit offsets can result in > >some pretty broken code. I've reproduced several problems on > >BGL/BGP/linux. The problems are most obvious with romio files > 2G or > >virtual addresses > 2G. > > Earlier this month I posted some comments to this mailing list about > 32-bit MPICH/ROMIO/MPI_Aint issues (see above and the archive). > > We've started to work on the problem. We're (optionally) making > MPI_Aint a 64 bit signed value even when integers/pointers are 32 bit > values. >

3 2

64-bit MPI_Aint/32-bit MPICH2/ROMIO
by Bob Cernohous 28 Feb '08

28 Feb '08

This is the initial patch for the work I described in my earlier email. It includes some configure changes to make MPI_Aint 8 bytes. It has some basic macros which we'll be using in the code to verify/test assumptions and to safely cast datatypes. And it has some initial code changes. I tried to separate out irrelevant changes in our local tree. It will be easier when our code base settles on an official MPICH2 release.

1 1

Announcing the availability of MPICH2-1.0.7rc1
by Pavan Balaji 15 Feb '08

15 Feb '08

Dear all, The first release candidate for the next release of MPICH2 (1.0.7) is available from http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/mpich2-1…. Please try it out and let us know if you encounter any problems. If all is well, we plan to release 1.0.7 in a week or so. The major changes since 1.0.6p1 are highlighted below. (Windows binaries are available from http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/mpich2-1… and http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/mpich2-1…) Regards, The MPICH2 Team --------------------------------------------------------------------- Main changes in 1.0.7 ===================== # PMI: Added support to use PMI Clique functionality for process managers that support it. # OVERALL: Added datatype free hooks to be used by devices independently. # OVERALL: Added device-specific timer support. # OVERALL: make uninstall works cleanly now, except for MPD installations. # ROMIO: Support to take hints from a config file # ROMIO: more tests and bug fixes for nonblocking I/O # SMPD: Singleton Init support. # WINDOWS: Fortran 90 support added. # SCTP: Added MPICH_SCTP_NAGLE_ON support. # MPE: Updated MPE logging API so that it is thread-safe (through global mutex). Updated all related documentation, e.g. man pages. # MPE: Added infrastructure to piggyback argument data to MPI states. # Many other bug fixes, memory leak fixes and code cleanup. A full list of changes is available using: svn log -r100:455 https://svn.mcs.anl.gov/repos/mpi/mpich2/trunk --------------------------------------------------------------------- -- Pavan Balaji http://www.mcs.anl.gov/~balaji

1 0

MPIEXEC, MPICH2 and getchar
by Ilkka Karasalo 01 Feb '08

01 Feb '08

I recently switched to MPICH2 after using MPICH for more than a decade. When running in interactive mode, my programs use the C functions getchar and putchar for character-per-character reading of keystrokes and writing characters to the display (standard output). This functionality is quite useful, enabling control of the interpretation of keystrokes, in particular keys generating escape sequences (such as e.g. the arrow keys). In programs started with MPIEXEC, however, the functionality of getchar is lost. Under MPIEXEC characters and character sequences generated keystrokes are read, written one-by-one to the display, and put into a buffer. The buffer is sent to the input stream of my programs getchar calls only at a keystroke (like the enter key) generating an end-of-line character. Thus, e.g., an arrow key results in the chacters of the generated escape sequence being written to the display, instead of being sent to the input stream of 'getchar'. This is, of course, quite annoying and destroys the intended interpretation of keystrokes. Any help with a work-around for this feature of MPIEXEC would be most welcome. Sincerely, Ilkka Karasalo -- Ilkka Karasalo FOI, Systems Technology KTH, MWL Gullfossgatan 6 Teknikringen 8 SE-164 90 Stockholm SE-100 44 Stockholm Phone: +46-8-55503627 Phone: +46-8-7908017 Fax: +46-8-55503869 Fax: +46-8-7906122 Email: ilkka.karasalo(a)foi.se Email: ilkkak(a)kth.se

2 1