Arnaud Giersch [Fri, 21 Dec 2012 12:56:27 +0000 (13:56 +0100)]
Reindent (again).
Martin Quinson [Tue, 8 Jan 2013 09:26:49 +0000 (10:26 +0100)]
new step to not forget when releasing the files
Maximiliano Geier [Mon, 7 Jan 2013 17:51:50 +0000 (18:51 +0100)]
chainsend: add optional argument to deployment file generator, so that we can generate a deployment file for a subset of the nodes in the platform file
Marion Guthmuller [Mon, 7 Jan 2013 12:56:08 +0000 (13:56 +0100)]
model-checker : comment backtrace and add ignore mechanism in mmalloc metadata
Marion Guthmuller [Thu, 3 Jan 2013 14:55:34 +0000 (15:55 +0100)]
model-checker : fix dichotomic search and insertion for visited states
Marion Guthmuller [Thu, 3 Jan 2013 11:13:12 +0000 (12:13 +0100)]
model-checker : update list of visited states during the exploration with the newest states
Marion Guthmuller [Wed, 2 Jan 2013 17:59:52 +0000 (18:59 +0100)]
model-checker : use xbt_abort after exhaustive exploration
Marion Guthmuller [Wed, 2 Jan 2013 17:47:08 +0000 (18:47 +0100)]
model-checker : ignore some local variables from simix in MC_init
Marion Guthmuller [Wed, 2 Jan 2013 17:38:17 +0000 (18:38 +0100)]
model-checker : fix sorting of visited states and state equality reduction
Marion Guthmuller [Tue, 1 Jan 2013 15:08:31 +0000 (16:08 +0100)]
model-checker : sorting of visited states according to chunks used
Marion Guthmuller [Tue, 1 Jan 2013 15:04:02 +0000 (16:04 +0100)]
model-checker : update bugged1 tesh
Martin Quinson [Fri, 28 Dec 2012 15:45:23 +0000 (16:45 +0100)]
Revert "Kill the <backbone> tag"
This reverts commit
7d248dfd0b7fa18bacf5403efbf3dcf7a0cf902f.
This is actually used in some situations where I don't have any good
replacement. Sorry for the noise.
Martin Quinson [Fri, 28 Dec 2012 15:29:02 +0000 (16:29 +0100)]
Kill the <backbone> tag
This was allowed only in the <cluster> tag, and was dupplicating the
features of the bb_* attributes of that tag. Kill this because the
parser is already complex enough without dupplicating ways to get a
given feature implemented. Sorry for the inconvinience.
Marion Guthmuller [Wed, 26 Dec 2012 14:37:03 +0000 (15:37 +0100)]
model-checker : update chord_neverjoin tesh
Marion Guthmuller [Wed, 26 Dec 2012 14:33:20 +0000 (15:33 +0100)]
model-checker : don't ignore simcall result in system state comparison
Martin Quinson [Wed, 26 Dec 2012 08:54:30 +0000 (09:54 +0100)]
add my last answer on the list to the FAQ
Martin Quinson [Tue, 25 Dec 2012 20:07:41 +0000 (21:07 +0100)]
this file should not overload the codebase
Marion Guthmuller [Tue, 25 Dec 2012 19:56:12 +0000 (20:56 +0100)]
model-checker : stats XP of model checking on chord example
Martin Quinson [Mon, 24 Dec 2012 22:17:08 +0000 (23:17 +0100)]
remove the 'new_' part of function name sg_platf_new_trace_connect
Martin Quinson [Mon, 24 Dec 2012 22:14:41 +0000 (23:14 +0100)]
acknoledge that the default value is set in network.c so that we have a single point of truth
Martin Quinson [Thu, 20 Dec 2012 15:22:05 +0000 (16:22 +0100)]
This set was recently renamed
suter [Fri, 21 Dec 2012 13:34:17 +0000 (14:34 +0100)]
first try at replaying alltoall. Not tested yet
suter [Fri, 21 Dec 2012 12:18:05 +0000 (13:18 +0100)]
make it compile
suter [Fri, 21 Dec 2012 11:37:36 +0000 (12:37 +0100)]
factor verbose logging
add dummy actions for unimplemented MPI calls needed by NPB codes
suter [Fri, 21 Dec 2012 10:30:41 +0000 (11:30 +0100)]
Cosmetics
Paul Bédaride [Fri, 21 Dec 2012 09:21:05 +0000 (10:21 +0100)]
Revert "make MSG_comm_waitany use dynar"
This reverts commit
ae811896a4746542e8de1616e8bc95a3ee6320f2.
Conflicts:
examples/msg/actions/actions.c
Arnaud Legrand [Thu, 20 Dec 2012 22:51:26 +0000 (23:51 +0100)]
Allow MSG one-side communications that rely on parallel task to work with the compound model.
Marion Guthmuller [Thu, 20 Dec 2012 22:18:16 +0000 (23:18 +0100)]
model-checker : fix MC_take_snapshot if malloc used for raw heap and several data segments for binary or libsimgrid
Marion Guthmuller [Thu, 20 Dec 2012 14:38:45 +0000 (15:38 +0100)]
model-checker : fix dpor with state equality reduction
Marion Guthmuller [Mon, 17 Dec 2012 10:47:21 +0000 (11:47 +0100)]
model-checker : if frame name is empty, stop stack unwinding
Marion Guthmuller [Mon, 17 Dec 2012 09:13:29 +0000 (10:13 +0100)]
model-checker : remove restriction of DPOR if several requests for same process
Maximiliano Geier [Thu, 20 Dec 2012 18:33:47 +0000 (19:33 +0100)]
chainsend: fix tesh
Maximiliano Geier [Thu, 20 Dec 2012 18:21:33 +0000 (19:21 +0100)]
chainsend: change peers active wait to waitall
Maximiliano Geier [Thu, 20 Dec 2012 17:32:33 +0000 (18:32 +0100)]
chainsend: broadcaster should send synchronously
Arnaud Giersch [Thu, 20 Dec 2012 17:07:16 +0000 (18:07 +0100)]
Now, you have to use "cursor" instead of "i".
Arnaud Giersch [Thu, 20 Dec 2012 17:05:42 +0000 (18:05 +0100)]
Reindent file before changes.
Arnaud Giersch [Thu, 20 Dec 2012 16:36:46 +0000 (17:36 +0100)]
Use "debug" log level here.
Arnaud Giersch [Thu, 20 Dec 2012 14:42:43 +0000 (15:42 +0100)]
Some symbols were renamed by last flexml update.
Arnaud Giersch [Thu, 20 Dec 2012 13:43:51 +0000 (14:43 +0100)]
Update with newer flexml.
Paul Bédaride [Thu, 20 Dec 2012 16:29:45 +0000 (17:29 +0100)]
make MSG_comm_waitany use dynar
Martin Quinson [Thu, 20 Dec 2012 15:17:07 +0000 (16:17 +0100)]
improve the doc to include MSG_config
Martin Quinson [Thu, 20 Dec 2012 15:08:57 +0000 (16:08 +0100)]
make MSG_config public, document it, and implement it properly, not through a crude hack
Paul Bédaride [Thu, 20 Dec 2012 13:29:48 +0000 (14:29 +0100)]
Move module tracing into corresponding modules
Arnaud Giersch [Thu, 20 Dec 2012 10:07:31 +0000 (11:07 +0100)]
Link with -ldl when using dlsym.
Martin Quinson [Thu, 20 Dec 2012 08:33:36 +0000 (09:33 +0100)]
remove the create_resource entry point in resource model APIs
That's a leftover from ancient times that is not used anymore.
Instead, these functions are registered by each model as callbacks
within the sg_platf mechanism, which is now the unique way to create
resources (we had up to 3 different ways to create resources in the
code base :)
Martin Quinson [Thu, 20 Dec 2012 08:21:02 +0000 (09:21 +0100)]
fix a typo
Martin Quinson [Thu, 20 Dec 2012 08:06:02 +0000 (09:06 +0100)]
rewrite the install doc, and integrate latest feedback on Mac builds
Martin Quinson [Thu, 20 Dec 2012 08:04:45 +0000 (09:04 +0100)]
mark as advanced some options that are useless to the users, and improve some doc
Martin Quinson [Thu, 20 Dec 2012 00:49:28 +0000 (01:49 +0100)]
remove a bit of cruft from the install doc, more remains to be cleaned
Martin Quinson [Wed, 19 Dec 2012 19:16:02 +0000 (20:16 +0100)]
cosmetics
Martin Quinson [Wed, 19 Dec 2012 18:41:07 +0000 (19:41 +0100)]
forgot the test file
Martin Quinson [Wed, 19 Dec 2012 18:31:22 +0000 (19:31 +0100)]
use system malloc for the raw heap if GNU ld is used
suter [Wed, 19 Dec 2012 11:18:52 +0000 (12:18 +0100)]
make this example be more valgrind compliant by increase host speed.
When tasks end after the "Finalize" reception, nothing is safe ...
Still has an invalid free in the end.
did some typo correction too
Martin Quinson [Tue, 18 Dec 2012 15:17:49 +0000 (16:17 +0100)]
Use xbt_malloc (and friends) instead of direct malloc()
This is good practice as it asserts that the result is not null, and
it actually help me right now when digging in the MC.
Arnaud Giersch [Tue, 18 Dec 2012 15:53:56 +0000 (16:53 +0100)]
Add -Wall to CMAKE_Fortran_FLAGS, unless enable_compile_warnings=OFF.
This triggers lots of warnings, mainly about implicit function declarations,
but I don't know how to fix them properly for now.
Arnaud Giersch [Tue, 18 Dec 2012 12:45:56 +0000 (13:45 +0100)]
Linker option is useless when '-c' is present.
suter [Tue, 18 Dec 2012 10:43:18 +0000 (11:43 +0100)]
this was fixed too
suter [Tue, 18 Dec 2012 10:10:58 +0000 (11:10 +0100)]
mention bug fix in ChangeLog
suter [Tue, 18 Dec 2012 09:54:18 +0000 (10:54 +0100)]
add test_sd_avail.tesh to the list
suter [Tue, 18 Dec 2012 09:50:50 +0000 (10:50 +0100)]
add a tesh file for the test of availability traces
Arnaud Giersch [Tue, 18 Dec 2012 09:21:45 +0000 (10:21 +0100)]
Avoid to compile files when they're not needed.
Arnaud Giersch [Tue, 18 Dec 2012 09:16:01 +0000 (10:16 +0100)]
This file does not belong to TRACING_SRC; define SIMGRID_SRC.
Arnaud Giersch [Tue, 18 Dec 2012 09:02:45 +0000 (10:02 +0100)]
Really remove deleted file.
Martin Quinson [Tue, 18 Dec 2012 09:20:26 +0000 (10:20 +0100)]
Let's be honnest
Martin Quinson [Mon, 17 Dec 2012 14:41:32 +0000 (15:41 +0100)]
refresh the README, and re-add it to the archives
Martin Quinson [Mon, 17 Dec 2012 14:41:10 +0000 (15:41 +0100)]
that was really ancient (shame shame)
suter [Tue, 18 Dec 2012 09:11:12 +0000 (10:11 +0100)]
This prevent periodic availability/state traces to loop forever
Paul Bédaride [Tue, 18 Dec 2012 09:01:10 +0000 (10:01 +0100)]
Fix _sg_do_model_check use with model_check=OFF
Paul Bédaride [Tue, 18 Dec 2012 08:28:16 +0000 (09:28 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
suter [Mon, 17 Dec 2012 22:43:34 +0000 (23:43 +0100)]
cosmetics
suter [Mon, 17 Dec 2012 22:40:23 +0000 (23:40 +0100)]
add an example to test availability traces. Weird behavior and runs
forever (thus no tesh yet)
Paul Bédaride [Mon, 17 Dec 2012 16:20:34 +0000 (17:20 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Paul Bédaride [Mon, 17 Dec 2012 16:20:27 +0000 (17:20 +0100)]
Restructure config headers
suter [Mon, 17 Dec 2012 15:33:27 +0000 (16:33 +0100)]
no need to explicitly tell that resources are initially up anymore.
suter [Mon, 17 Dec 2012 15:32:20 +0000 (16:32 +0100)]
when the first event of a trace (availability or state) starts at a time
greater than 0, this means that the default value has to be used
meanwhile. Create a dummy event whose value is -1 and handle this in
surf_presolve.
suter [Mon, 17 Dec 2012 15:30:12 +0000 (16:30 +0100)]
when the value of an event is a negative one, don't update resource
state during presolve. Allows to use default status up to the time (>0)
of the "real" first event.
suter [Mon, 17 Dec 2012 14:42:08 +0000 (15:42 +0100)]
add more debugging info there. Nice to know what is the new value
assigned to the link
Maximiliano Geier [Mon, 17 Dec 2012 14:52:20 +0000 (15:52 +0100)]
chainsend: print some transfer stats, move debug prints to where they belong
also count up bytes when transfering, so that the peer can print its "transfer speed" when it's done, and finally update tesh accordingly
Paul Bédaride [Mon, 17 Dec 2012 10:17:04 +0000 (11:17 +0100)]
Fix model checker config issue
Paul Bédaride [Mon, 17 Dec 2012 10:04:19 +0000 (11:04 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Paul Bédaride [Mon, 17 Dec 2012 10:03:42 +0000 (11:03 +0100)]
Move config from surf to simgrid
Sebastien Badia [Sun, 16 Dec 2012 23:46:49 +0000 (00:46 +0100)]
Cosmetics : Remove whitespaces for mquinson :-)
Marion Guthmuller [Sun, 16 Dec 2012 20:18:01 +0000 (21:18 +0100)]
model-checker : remove tesh bugged2 (too long with the new version of DPOR)
Marion Guthmuller [Sun, 16 Dec 2012 20:14:25 +0000 (21:14 +0100)]
model-checker : update tesh bugged1 with the new version of DPOR
Marion Guthmuller [Sun, 16 Dec 2012 17:31:45 +0000 (18:31 +0100)]
model-checker : interleave all enabled processes for each state and apply DPOR only once per process
Marion Guthmuller [Fri, 14 Dec 2012 15:24:09 +0000 (16:24 +0100)]
model-checker : cosmectics
Marion Guthmuller [Fri, 14 Dec 2012 14:57:57 +0000 (15:57 +0100)]
Remove comm from src_proc->comms and dst_proc->comms when finish it,
if is detached (dsend)
Currently, with dsend action, the wait of src_proc is not necessary to
finish the communication. In this case, only dst_proc removes this
communication from its list and destroys it, if src_proc is always
running. When src_proc makes a process_cleanup after that, it
retrieves this communication in its list while it doesn't exist
anymore and thinks that this communication isn't for it. This led to a
simulation error.
Marion Guthmuller [Fri, 14 Dec 2012 14:50:35 +0000 (15:50 +0100)]
remove comm from rdv if comm destroyed before being finished
Martin Quinson [Fri, 14 Dec 2012 15:02:05 +0000 (16:02 +0100)]
doc fix: AS has no 'name' attribute, but a 'id' one
Martin Quinson [Fri, 14 Dec 2012 14:51:43 +0000 (15:51 +0100)]
reindent, no real change
Arnaud Giersch [Fri, 14 Dec 2012 12:39:22 +0000 (13:39 +0100)]
Typos.
Arnaud Giersch [Fri, 14 Dec 2012 12:39:22 +0000 (13:39 +0100)]
Use cat+here document, instead of non-portable echo -e.
Martin Quinson [Fri, 14 Dec 2012 11:12:42 +0000 (12:12 +0100)]
make doxygen happy
Martin Quinson [Thu, 13 Dec 2012 15:36:24 +0000 (16:36 +0100)]
SimGrid 3.7 is ways too recent for GRAS (+cosmetics)
Martin Quinson [Thu, 13 Dec 2012 15:27:01 +0000 (16:27 +0100)]
document how to react when stuff break during the upgrade
Martin Quinson [Tue, 11 Dec 2012 10:52:30 +0000 (11:52 +0100)]
ignore that binary file
Augustin Degomme [Wed, 12 Dec 2012 18:33:40 +0000 (19:33 +0100)]
change the way we handle MPI_Request termination
to avoid using smx_action too much in smpi.
Removes a few more leaks in collectives
Handling of waitany was changed, it should return sooner and avoid a few unnecessary simcalls
Arnaud Legrand [Wed, 12 Dec 2012 11:00:35 +0000 (12:00 +0100)]
Spell check...
Arnaud Legrand [Wed, 14 Nov 2012 21:55:02 +0000 (14:55 -0700)]
This is useful for those typing ./configure & make without looking at output. :)
Paul Bédaride [Wed, 12 Dec 2012 09:11:54 +0000 (10:11 +0100)]
Add ctest2junit.xsl for jenkins