Change Summary for MPICH2 for week of Mar 25, 2007

Information on all weekly change summaries is available.

List of revisions

• (gropp) Added ABI Version information to channels
• (gropp) Added ERR_SPAWN to known errors list
• (gropp) Added missing segment free
• (rross) All the dataloop code in the ROMIO tree now builds. Will bring changes backto MPICH2 side next and verify correctness before beginning to use in ROMIO.
• (rross) Being more careful about #includes.
• (rross) cleaned up some prototypes, etc.
• (rross) compile errors :).
• (rross) continuing to do basic work to get things together. still haven't changedtop-level configure/make to enable build of this code in HEAD.
• (rross) Copied back from ROMIO side.
• (rross) copied back from ROMIO side.
• (rross) file that should have to change when moving from place to place is dataloop.h.
• (rross) final checkin of dataloop changes from the ROMIO side.
• (balaji) First draft of changes. Need to debug.
• (rbutler) Fix minor bug related to ifhns when using -machinefile option.
• (rross) fix to hack.
• (balaji) Fixed syntax errors.
• (rross) for use in other contexts. This makes it easy to move it around. The onlyfile that should have to change when moving from place to place is dataloop.h.
• (rross) getting closer to compiling...still not checking in makefile etc. so as tonot break everything.
• (robl) John Bent finds a bug with collective buffering of non-interleaved accesses
• (rross) keep things tidy.
• (balaji) More synchronization
• (rross) Move assert() calls to DLOOP_Assert() calls.
• (balaji) Moved around functions to avoid global variables.
• (rross) Moving the dataloop implementation into its own subdirectory in preparationfor use in other contexts. This makes it easy to move it around. The onlyfile that should have to change when moving from place to place is dataloop.h.
• (buntinas) nemesis blocking: added buggy shm blocking code
• (buntinas) nemesis blocking: added macros to disable blocking
• (buntinas) nemesis blocking: added polling loop before sleep
• (buntinas) nemesis blocking: bugfix -- process is now correctly woken when a cell is returned to its free queue
• (buntinas) nemesis blocking: bugfix in blocking vs nonblocking progress
• (buntinas) nemesis blocking: bugfix to signal process when free cells are returned to an empty queue
• (buntinas) nemesis blocking: debugging
• (buntinas) nemesis blocking: errcode fix
• (buntinas) nemesis blocking: fixed alignment issues in nem_queue struct
• (buntinas) nemesis blocking: fixed deadlock in blocking shm
• (buntinas) nemesis blocking: making newtcp netmod threaded
• (buntinas) nemesis blocking: thread bugfixes
• (buntinas) nemesis: fixed nemesis to use segment_ptr
• (buntinas) nemesis: removed nemesis packet types from CH3 packet structure
• (rross) new subdir to work with (dataloop).
• (rross) not break everything.
• (rross) rough, uncompiled version of dataloops in romio. no makefile yet, so nocompile errors :).
• (rross) Shifting around dataloop code to make it easier to use in other projects.
• (balaji) Still buggy; just synchronizing.
• (balaji) Synchronizing with Darius.
• (rross) temporary hack to work around objects with same name conflicting in library.
• (rross) to MPICH2 side next and verify correctness before beginning to use in ROMIO.
• (rross) top-level configure/make to enable build of this code in HEAD.
• (rbutler) Updates to handle "new" messages to/from pmi_simple whichwere added by Bill Gropp.
• (gropp) Use $(MAKE) instead of make and all clean-local target for the subchannels

Files for each revision