Change Summary for MPICH2 for week of Aug 12, 2007

Information on all weekly change summaries is available.

List of revisions

• (robl) (PVFS v2, zoid) can do a 'scalable open'
• (toonen) a small fix to allow MPICH2 to be compiled with insure++
• (rbutler) Add try block around send_dict_msg
• (toonen) added @..._OTHER_LIBS@ to MPI_OTHERLIBS in the Fortran conf files to make them consistent with the compiler and bash script files
• (toonen) added a couple of FUNCNAME definitions
• (toonen) added a simple hello world program for Fortran 77
• (toonen) added an option to force MPICH2 to call abort() when an error first occurs. calling abort() allows the OS the opportunity to produce a core file which the developer can use to examine the state prior to the error occurring
• (toonen) added check for the presence of the Fortran 2003 intrinsic routines get_command_argument and command_argument_count
• (toonen) added code to determine if and how to extract command line arguments from Fortran programs. also exported Fortran to C mappings so that the device's configure can access the information.
• (toonen) added code to rename MPIX and PMIX symbols
• (toonen) added code to the Fortran MPI_INIT and MPI_INIT_THREAD routines to extract command line arguments and pass them to their C counterparts. the globus device needs this since some vendor MPI implementations require the command line arguments (like MPICH-GM)
• (toonen) added device hooks for the various status definition macros. also added documentation for the MPID_DEV_COMM_FUNC_HOOK macro.
• (toonen) added hook macros that allow the device to track the creation and destruction of datatypes
• (toonen) added hooks for devices that need to perform detailed tracking of communicator operations
• (chan) added memset() to initialize CLOG_CommIDs_t array after malloc and realloc to keep valgrind happy on 64bit machine.
• (balaji) Adding optimization support for the IBM xlc/xlf compilers.
• (rbutler) Chg if to elif in handling exceptions during send_dict_msg
• (rbutler) Chg pass to break in handling PIPE errs in send_dict_msg
• (robl) declare and use feature test in UFS (common)
• (jayesh) Due to potential blocking of MPIDI_PG_GetConnString() the state of VC is now *_CONNECTING in MPIDU_Sock_connect(). Queueing requests when VC is in *_CONNECTING state. Changing state of VC to *_CONNECTING before invoking MPIDU_Sock_connect() from within ssm
• (balaji) Enable -O2 optimization by default. This optimizes, gcc, intel, Portland, Sun and Absoft compilers. More optimization flags might need to be added for other compilers.
• (robl) fix the statvfs f_basetype check for solaris
• (balaji) Fixed a compilation error.
• (toonen) fixed a couple of errors that caused the tail end of strings not to be parsed correctly. also added casts to eliminated warnings.
• (balaji) Fixed a typo in setting the socket state.
• (balaji) Include IPC only when USE_SINGLE_MSG_QUEUE is defined.
• (balaji) Increased the number of iterations to reduce noise.
• (toonen) modified the code that generates dependency targets so that a separate -MT option is used for each object file. this allows compiler scripts that do not get quoting quite right to still generate dependency information.
• (robl) move the feature test for shared file pointers down a bit: feature tests don't work until we hvae a valid ADIO_File
• (toonen) moved the declarations of MPI_F_STATUS_IGNORE and MPI_F_STATUSES_IGNORE to a point after the inclusion of the device's header file. this allow the device to rename these symbols should it chose to do so.
• (buntinas) nemesis: fixed hang bug in nemesis progress engine. Added code in blocking receive progress to check whether network module completed a request.
• (robl) one more declaration
• (toonen) oops...reversing commit that was supposed to be applied to the mpig branch
• (robl) optimizations, others (Lustre, UFS) can have one process create, and others(PVFS v2, zoid) can do a 'scalable open'
• (balaji) Protecting header includes in a macro. Otherwise, they give errors on windows.
• (jayesh) Recursive spawns fail since the search path of the exec for the spawned process is set to the incorrect working directory. Setting the working directory of an MPI process to current working directory if not already set.
• (robl) remove unused variables
• (jayesh) Removed exporting CLOG_CommIDs4World
• (toonen) removed sr and srvec from the list of extra programs. these program were moved to test/basic some time ago.
• (toonen) removed the globus device from the list of devices to remove from the distribution tarball
• (chan) restored CLOG_COMM_TABLE_INCRE to the original default value
• (robl) rework some of the ADIO_Open logic. some file systems (NFS) cannot do anyoptimizations, others (Lustre, UFS) can have one process create, and others(PVFS v2, zoid) can do a 'scalable open'
• (toonen) send usage statistics to default server rather than Mike's personal test server
• (chan) split the MPE section into subsections that describe MPI automatic and MPE user-defined logging.
• (robl) sync up with HEAD
• (robl) the MPIU macros aren't available when building standalone
• (balaji) The test was doing Testall a second time even if the first testall is successful. This is incorrect according to the MPI standard, since the requests are set to MPI_REQUEST_NULL as soon as the first Testall is successful.
• (toonen) undefine any existing MPID_Wtick macro before redefining it to use the generic implementation. this is needed by the globus device which may have already defined MPID_Wtick as part of its internal symbol renaming process.
• (balaji) Updated changes file to mention the compiler optimizations support.
• (chan) updated comment in sample code.
• (toonen) updated notes to reflect items that need completing
• (chan) updated version info.
• (toonen) updates to reflect the changes to the MPICH2 framework since 1.0.4p1
• (robl) use the new names for feature test flags/functions
• (jayesh) With SMPD, MPIDI_PG_GetConnString() blocks on the progress engine allowing another thread to post a connect for a connection in the CONNECTING state. Changed the state of connection to CONNECTING before blocking on MPIDI_PG_GetConnString() & queued data when we find connection in the CONNECTING state

Files for each revision