- 06 Nov, 2008 4 commits
-
-
Pavan Balaji authored
[svn-r3456] Getting rid of SVN IDs that are present in 5% of the files and not in the other 95%. This is creating unnecessary conflicts for collaborators who merge from MPICH2.
-
Rajeev Thakur authored
-
Rajeev Thakur authored
-
David Goodell authored
Also add an implementation based on Brice's knem module. This change also fixes several minor bugs in the LMT code. Reviewed by buntinas@.
-
- 04 Nov, 2008 1 commit
-
-
Rob Latham authored
-
- 03 Nov, 2008 2 commits
-
-
Pavan Balaji authored
[svn-r3442] Removed the asynchronous SIGCHLD handler and made the code completely synchronous. This needs to be tested on other platforms for correctness.
-
Pavan Balaji authored
[svn-r3441] OR the exit status of each process and return it to the user. This works well when all the exit codes are the same (e.g., all of them completed successfully or all of them failed with the same error code). But when different processes return different error codes, their OR'ed value might not match any one of them.
-
- 02 Nov, 2008 2 commits
-
-
Pavan Balaji authored
[svn-r3440] Several fixes and cleanups to the bootstrap signal handling. The return status is still not correct, but it's not clear what the return status should be.
-
Pavan Balaji authored
[svn-r3439] Fixed some more memory leaks and moved non-bootstrap server specific code to utility functions.
-
- 01 Nov, 2008 6 commits
-
-
Pavan Balaji authored
[svn-r3438] Some compilers don't seem to like stdin, stdout and stderr var names too much even if they are within a structure.
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
[svn-r3434] White space change using the cleanup script at http://www.mcs.anl.gov/~balaji/personal/code-cleanup.sh
-
Pavan Balaji authored
-
- 31 Oct, 2008 1 commit
-
-
Darius Buntinas authored
[svn-r3432] Fixed #213 -- mg failures. The problem was that if side A completed an LMT, freed the request, then allocated a new request (with the same handle) and started another LMT before side B finished the first LMT, side A's new request could match the old LMT transfer and use the same (old) buffer. This was fixed by clearing the remote_req_id field as soon as both sides agreed on which request to transfer.
-
- 30 Oct, 2008 18 commits
-
-
Pavan Balaji authored
-
Pavan Balaji authored
[svn-r3430] Generalized some of the environment manipulation functions as utils; a lot more needs to be moved out of specific components.
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
[svn-r3424] Fixed a memory leak in the stdin fd; this fix is correct, but it seems to have showed up a number of other problems now.
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Jayesh Krishna authored
[svn-r3417] 1) Support MPI C types in FORTRAN (2) Enable SMP collectives & add src to map proc to localhost (3) Enable debugging handle allocs, Use 16-bits for rank in pkt header, Allow use of __FUNCTION__, No valgrind headers in windows
-
Pavan Balaji authored
[svn-r3416] Disable X by default even if it's set in the user's settings; it's causing too many problems.
-
Pavan Balaji authored
-
Pavan Balaji authored
[svn-r3414] Allow all three modes for X forwarding -- explicitly allow, explicitly disallow or do whatever is set in the user's configuration.
-
Pavan Balaji authored
-
Pavan Balaji authored
-
- 29 Oct, 2008 6 commits
-
-
Pavan Balaji authored
-
Pavan Balaji authored
-
Pavan Balaji authored
[svn-r3409] 1. Allow reuse of hostfile from the previous executable if a multi-executable command-line does not specify one for some executables. 2. Several bug fixes in returning errors to the upper layer.
-
William Gropp authored
[svn-r3408] Removed misplaced thread critical sections in an internal routine (the ALLFUNC critical section must be used only in the top-level MPI routines) and added a missing critical section exit
-
William Gropp authored
[svn-r3407] Removed misplaced thread critical sections in an internal routine (the ALLFUNC critical section must be used only in the top-level MPI routines) and added a missing critical section exit
-
Darius Buntinas authored
[svn-r3406] fixed bug in lmt code: When a cookie is sent along with the RTS, and the request is matched immediately, we weren't calling do_cts. This wasn't caught in our tests because none of the existing netmods or shm send cookies with an RTS. This implies that the new code still hasn't been tested, but looks better than before. Reviewed by Dave.
-