Augustin Degomme [Thu, 6 Dec 2012 17:43:06 +0000 (18:43 +0100)]
handle proper cleanup of mpi datatypes with MPI_Type_free with a refcounter
Augustin Degomme [Thu, 6 Dec 2012 17:31:11 +0000 (18:31 +0100)]
mallocators were not initialized and used since commit
1f50f809
Paul Bédaride [Thu, 6 Dec 2012 09:57:01 +0000 (10:57 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Paul Bédaride [Thu, 6 Dec 2012 09:52:20 +0000 (10:52 +0100)]
Add ModelChecker build type
Marion Guthmuller [Thu, 6 Dec 2012 09:31:51 +0000 (10:31 +0100)]
model-checker : liveness model-checking works only on 64bits for now, add condition for tests
Marion Guthmuller [Thu, 6 Dec 2012 09:16:53 +0000 (10:16 +0100)]
model-checker : improve debug message if different global variable
Marion Guthmuller [Thu, 6 Dec 2012 09:16:10 +0000 (10:16 +0100)]
model-checker : ignore global variables corresponding to a counter for coverage
Martin Quinson [Wed, 5 Dec 2012 20:57:59 +0000 (21:57 +0100)]
these includes are useless
Martin Quinson [Wed, 5 Dec 2012 20:21:18 +0000 (21:21 +0100)]
we need that files to compile on windows
Augustin Degomme [Wed, 5 Dec 2012 18:10:42 +0000 (19:10 +0100)]
activate pt2pt truncmult test in teshsuite
Augustin Degomme [Wed, 5 Dec 2012 17:55:18 +0000 (18:55 +0100)]
modify the way we handle MPI_ANY_SOURCE, MPI_ANY_TAG, and MPI_TRUNCATED.
Fix the case where a request may live outside smpi and be reused
We store the values we may change in other fields
Augustin Degomme [Wed, 5 Dec 2012 17:51:24 +0000 (18:51 +0100)]
make valgrind happy with smpi by handling action destruction inside finish_wait
Marion Guthmuller [Wed, 5 Dec 2012 15:21:15 +0000 (16:21 +0100)]
model-checker : increase tesh timeout for mc-chord-neverjoin-ucontext
Marion Guthmuller [Wed, 5 Dec 2012 14:58:31 +0000 (15:58 +0100)]
model-checker : update snapshot_comparison tests
Marion Guthmuller [Tue, 4 Dec 2012 18:16:33 +0000 (19:16 +0100)]
model-checker : ignore new variables
Marion Guthmuller [Tue, 4 Dec 2012 18:15:13 +0000 (19:15 +0100)]
model-checker : display instruction pointer if different local variable
Marion Guthmuller [Tue, 4 Dec 2012 15:30:28 +0000 (16:30 +0100)]
model-checker : ignore simcall result in heap comparison
Marion Guthmuller [Tue, 4 Dec 2012 15:29:13 +0000 (16:29 +0100)]
model-checker : remove unused code
Marion Guthmuller [Tue, 4 Dec 2012 15:28:54 +0000 (16:28 +0100)]
model-checker : set raw heap for initialization step
Marion Guthmuller [Tue, 4 Dec 2012 15:20:39 +0000 (16:20 +0100)]
model-checker : get libsimgrid_path with MC_init_memory_map_infos
bedaride [Tue, 4 Dec 2012 15:15:31 +0000 (16:15 +0100)]
Dirty Fix to handle model checking with new simcalls
Martin Quinson [Tue, 4 Dec 2012 13:19:52 +0000 (14:19 +0100)]
and now, with the right URL for the faq, bummer
Martin Quinson [Tue, 4 Dec 2012 13:16:05 +0000 (14:16 +0100)]
fix a bunch of links to the FAQ -- hate that code dupplication
Arnaud Giersch [Tue, 4 Dec 2012 12:13:30 +0000 (13:13 +0100)]
Avoid to add an empty component into LD_LIBRARY_PATH if it was not set before.
Also remove the trailing filename.
Marion Guthmuller [Mon, 3 Dec 2012 18:07:15 +0000 (19:07 +0100)]
model-checker : fix MC_state_set_executed_request
Comm for SIMCALL_COMM_WAIT was internal_comm address of mc_state
(confusion between s_smx_action_t and smx_action_t)
shenshei [Mon, 3 Dec 2012 16:54:02 +0000 (17:54 +0100)]
Fix debug message in simix
Augustin Degomme [Mon, 3 Dec 2012 16:01:19 +0000 (17:01 +0100)]
comment a bit of code in sender_gap for smpi, which leaked memory.
This is temporary to correct the leak, waiting for the replacement of sender_gap
Augustin Degomme [Mon, 3 Dec 2012 16:00:37 +0000 (17:00 +0100)]
remove sender's side data from mpi_send in the receiver's side for detached send
fixes a leak.
Augustin Degomme [Mon, 3 Dec 2012 11:10:47 +0000 (12:10 +0100)]
hopefully remove leak in permanent receive mode
Arnaud Giersch [Mon, 3 Dec 2012 09:29:13 +0000 (10:29 +0100)]
Add a note about inline functions in README.coding.
Arnaud Giersch [Fri, 30 Nov 2012 16:39:21 +0000 (17:39 +0100)]
Remove spurious semicolon.
Arnaud Giersch [Fri, 30 Nov 2012 16:31:16 +0000 (17:31 +0100)]
Define inline functions in header file (again).
Arnaud Giersch [Fri, 30 Nov 2012 16:30:56 +0000 (17:30 +0100)]
Remove prototypes.
These inline functions are only used in smx_user. Forward declarations are useless.
Arnaud Giersch [Fri, 30 Nov 2012 15:41:11 +0000 (16:41 +0100)]
Define inline function in header file.
Arnaud Giersch [Fri, 30 Nov 2012 15:23:18 +0000 (16:23 +0100)]
Initialize variables, and make valgrind happy.
Arnaud Giersch [Fri, 30 Nov 2012 14:59:14 +0000 (15:59 +0100)]
Print a message and exit if file was not found.
Arnaud Giersch [Fri, 30 Nov 2012 14:56:11 +0000 (15:56 +0100)]
Activate memcheck tests for smpi-replay.
Arnaud Giersch [Mon, 3 Dec 2012 09:53:27 +0000 (10:53 +0100)]
Make generate_memcheck_tests.pl handle command "mkfile".
Arnaud Giersch [Fri, 30 Nov 2012 14:54:31 +0000 (15:54 +0100)]
Set valgrind option --track-origins=no, to make it use less memory.
Actually, for the test smpi-replay, it needs more than 4GB of memory
with option --track-origins=yes.
Arnaud Giersch [Thu, 29 Nov 2012 21:20:30 +0000 (22:20 +0100)]
Make tests work when building out of the sources.
Arnaud Giersch [Thu, 29 Nov 2012 21:19:29 +0000 (22:19 +0100)]
Don't ignore full directory contents.
Marion Guthmuller [Sun, 2 Dec 2012 21:28:01 +0000 (22:28 +0100)]
model-checker : remove independance between iRecv and Wait
Marion Guthmuller [Sun, 2 Dec 2012 21:15:23 +0000 (22:15 +0100)]
model-checker : statistics printed twice if exhaustive exploration
Marion Guthmuller [Sun, 2 Dec 2012 21:13:37 +0000 (22:13 +0100)]
model-checker : get global variables with objdump and compare only the bytes in data segments corresponding to global variables
Marion Guthmuller [Sun, 2 Dec 2012 21:09:09 +0000 (22:09 +0100)]
model-checker : get segments start only in MC_init_memory_map_info
Marion Guthmuller [Sun, 2 Dec 2012 21:07:32 +0000 (22:07 +0100)]
model-checker : text segment is read and exec only
Marion Guthmuller [Sun, 2 Dec 2012 21:05:52 +0000 (22:05 +0100)]
model-checker : cleanups in comments
Marion Guthmuller [Sun, 2 Dec 2012 21:01:39 +0000 (22:01 +0100)]
model-checker : get .got.plt sections
Marion Guthmuller [Thu, 29 Nov 2012 13:09:04 +0000 (14:09 +0100)]
model-checker : increase tesh timeout
Marion Guthmuller [Thu, 29 Nov 2012 13:07:21 +0000 (14:07 +0100)]
model-checker : fix tesh name
Marion Guthmuller [Thu, 29 Nov 2012 13:02:15 +0000 (14:02 +0100)]
model-checker : increase tesh timeouts
suter [Sun, 2 Dec 2012 11:45:22 +0000 (12:45 +0100)]
add the test of allReduce replay
suter [Sun, 2 Dec 2012 11:44:30 +0000 (12:44 +0100)]
yet another action file for allReduce
suter [Fri, 30 Nov 2012 15:20:10 +0000 (16:20 +0100)]
missing file
suter [Fri, 30 Nov 2012 14:38:41 +0000 (15:38 +0100)]
hey you, you're supposed to be removed already (did I say that I hate
git when it come to rm already?)
suter [Fri, 30 Nov 2012 09:16:52 +0000 (10:16 +0100)]
remove temporary file at the end of the tesh f
suter [Fri, 30 Nov 2012 09:15:36 +0000 (10:15 +0100)]
typos
Augustin Degomme [Fri, 30 Nov 2012 14:59:39 +0000 (15:59 +0100)]
have the smpif90 file replace the main program clause by our own for f90
Augustin Degomme [Fri, 30 Nov 2012 11:44:15 +0000 (12:44 +0100)]
fix a memory leak happening with smpi_replay
Augustin Degomme [Wed, 28 Nov 2012 17:42:49 +0000 (18:42 +0100)]
add support of smpif90 file
Augustin Degomme [Wed, 28 Nov 2012 10:36:42 +0000 (11:36 +0100)]
few fortran defines to make bigdft happy
Augustin Degomme [Wed, 28 Nov 2012 10:35:49 +0000 (11:35 +0100)]
add yet another stunt with smpi_simulated_main to provide support for f90
Augustin Degomme [Wed, 28 Nov 2012 09:01:29 +0000 (10:01 +0100)]
replace smpi_simulated main by smpi_simulated_main__, because __ is wanted by fortran compilers
C compilers don't really care how we name it.
Augustin Degomme [Wed, 28 Nov 2012 08:58:41 +0000 (09:58 +0100)]
add two functions to fortran bindings, needed by specfem
Augustin Degomme [Wed, 28 Nov 2012 08:57:01 +0000 (09:57 +0100)]
remove unsupported characters from mpif.h, for f90 compiler to be happy
shenshei [Fri, 30 Nov 2012 14:09:01 +0000 (15:09 +0100)]
Merge ../simgrid
shenshei [Fri, 30 Nov 2012 14:07:38 +0000 (15:07 +0100)]
New simcalls work
Maximiliano Geier [Fri, 30 Nov 2012 10:30:58 +0000 (11:30 +0100)]
chainsend: fix compilation errors
Signed-off-by: Maximiliano Geier <maximiliano.geier@loria.fr>
Maximiliano Geier [Fri, 30 Nov 2012 10:11:05 +0000 (11:11 +0100)]
chainsend: change message sizes to more realistic values, avoid passing too many parameters around, use MSG_task_isend in broadcaster instead of send
Signed-off-by: Maximiliano Geier <maximiliano.geier@loria.fr>
suter [Fri, 30 Nov 2012 08:00:11 +0000 (09:00 +0100)]
forgot to add this file thus breaking the smpi_replay test
suter [Thu, 29 Nov 2012 16:13:40 +0000 (17:13 +0100)]
rm this file, generated now
suter [Thu, 29 Nov 2012 16:13:10 +0000 (17:13 +0100)]
add 2 more test for smpi replay: barrier and some Isend/Irecv
suter [Thu, 29 Nov 2012 16:06:50 +0000 (17:06 +0100)]
printing the final simulation time also has to be done by the last
process to finish.
smpi_process_finalize/destroy already called by PMPI_finalize (remove)
use my own global to count the number of active processes (to not mess
with process_count)
suter [Thu, 29 Nov 2012 15:40:26 +0000 (16:40 +0100)]
assuming that rank 0 is the last to finish is a bad idea. the last
process to die has to free the dicts.
suter [Thu, 29 Nov 2012 15:34:05 +0000 (16:34 +0100)]
the comm_size action is useless now
suter [Thu, 29 Nov 2012 15:27:09 +0000 (16:27 +0100)]
generate the one_trace file for each test (will allow me to test several
basic traces with a single temporary file)
Arnaud Giersch [Thu, 29 Nov 2012 13:37:37 +0000 (14:37 +0100)]
Enable memcheck for smpi tests too.
Tests from teshsuite/smpi/mpich-test/ are however still excluded, since
they are run through intricate scripts.
Also exclude smpi-replay, since it's too memory hungry.
Arnaud Giersch [Thu, 29 Nov 2012 13:37:22 +0000 (14:37 +0100)]
Make valgrind skip commands from /usr/bin/ and /bin/.
Arnaud Giersch [Thu, 29 Nov 2012 14:18:38 +0000 (15:18 +0100)]
Cosmetics: test ucontexts before raw contexts.
Arnaud Giersch [Thu, 29 Nov 2012 14:17:46 +0000 (15:17 +0100)]
Add missing guards.
suter [Thu, 29 Nov 2012 13:24:09 +0000 (14:24 +0100)]
remove some leaks by calling SD_exit() before dying
suter [Thu, 29 Nov 2012 12:35:48 +0000 (13:35 +0100)]
clean stuff even when the scheduled is not good
Arnaud Giersch [Wed, 28 Nov 2012 22:40:35 +0000 (23:40 +0100)]
Fix printf statements.
Arnaud Giersch [Wed, 28 Nov 2012 21:39:53 +0000 (22:39 +0100)]
Rename file, and make test smpi-mpich-pt2pt-raw succeed.
Arnaud Giersch [Wed, 28 Nov 2012 20:34:51 +0000 (21:34 +0100)]
Add a few thoughts in README.coding, and correct typos.
Arnaud Giersch [Wed, 28 Nov 2012 17:22:19 +0000 (18:22 +0100)]
We have status, return its value.
Arnaud Giersch [Wed, 28 Nov 2012 17:19:17 +0000 (18:19 +0100)]
Define functions as static.
Fix build errors from -Werror=missing-prototypes.
Arnaud Giersch [Wed, 28 Nov 2012 17:18:19 +0000 (18:18 +0100)]
Remove unused variables.
Arnaud Giersch [Wed, 28 Nov 2012 16:48:37 +0000 (17:48 +0100)]
Add missing prototypes.
Arnaud Giersch [Wed, 28 Nov 2012 17:14:30 +0000 (18:14 +0100)]
An inline function must be visible from header file.
Arnaud Giersch [Wed, 28 Nov 2012 16:47:43 +0000 (17:47 +0100)]
Return new value.
Did it work before?
Maximiliano Geier [Wed, 28 Nov 2012 16:26:03 +0000 (17:26 +0100)]
Oops, forgot to actually add the new tesh
Maximiliano Geier [Wed, 28 Nov 2012 16:06:55 +0000 (17:06 +0100)]
Remove unused tesh (copied from sendrecv), add a real tesh for chainsend
Maximiliano Geier [Wed, 28 Nov 2012 14:19:57 +0000 (15:19 +0100)]
Oops, forgot to change one XBT_INFO
Maximiliano Geier [Wed, 28 Nov 2012 14:10:27 +0000 (15:10 +0100)]
Some more cleaning up, commented out INFO logs go to DEBUG
Maximiliano Geier [Wed, 28 Nov 2012 12:31:20 +0000 (13:31 +0100)]
Code cleanup
Maximiliano Geier [Wed, 28 Nov 2012 11:01:54 +0000 (12:01 +0100)]
Fix compilation error: there's no MSG_clean anymore
Maximiliano Geier [Tue, 16 Oct 2012 12:07:36 +0000 (14:07 +0200)]
Fix typo
Maximiliano Geier [Tue, 16 Oct 2012 09:42:50 +0000 (11:42 +0200)]
Renamed project to chainsend to avoid confusion
Maximiliano Geier [Mon, 15 Oct 2012 16:14:01 +0000 (18:14 +0200)]
Deployment file generator in Ruby, takes a platform file, parses hostnames and outputs a deployment file for kadeploy Fixed mailbox names so that they don't depend on the hostnames FIXME: comm_t should be queued both in sends and recvs