unknown [Wed, 11 Jan 2012 10:06:44 +0000 (11:06 +0100)]
Add missing headers to simgrid install for windows
unknown [Wed, 11 Jan 2012 09:43:47 +0000 (10:43 +0100)]
New logo for installer.
unknown [Tue, 10 Jan 2012 21:42:29 +0000 (22:42 +0100)]
return 0 instead of die.
unknown [Tue, 10 Jan 2012 21:38:11 +0000 (22:38 +0100)]
On windows pthread_atfork is not implemented
unknown [Tue, 10 Jan 2012 21:37:49 +0000 (22:37 +0100)]
Make Hello World example works with mingw
unknown [Tue, 10 Jan 2012 21:16:32 +0000 (22:16 +0100)]
Update nsis script.
Update version of pcre lib.
unknown [Tue, 10 Jan 2012 20:42:10 +0000 (21:42 +0100)]
Don't compile those examples on Win
unknown [Tue, 10 Jan 2012 17:46:28 +0000 (18:46 +0100)]
Update README from HelloWorld.
Replace gmake by mingw32-make
unknown [Tue, 10 Jan 2012 17:38:00 +0000 (18:38 +0100)]
Update doc for MinGW
unknown [Tue, 10 Jan 2012 17:37:40 +0000 (18:37 +0100)]
sleep doesn't exist on windows.
unknown [Tue, 10 Jan 2012 17:14:29 +0000 (18:14 +0100)]
Cosmetics and replace some deprecated function 'exec_program' for 'execute_process'.
Remove unused part for windows.
unknown [Tue, 10 Jan 2012 17:05:36 +0000 (18:05 +0100)]
Make Simgrid compile with MinGW and gcc4.6
Need new flags.
For now on gcc 64bits it fails so we must use gcc 32bits on 64platforms.
unknown [Tue, 10 Jan 2012 17:03:57 +0000 (18:03 +0100)]
Remove unused part.
unknown [Tue, 10 Jan 2012 14:27:38 +0000 (15:27 +0100)]
Only use MinGW compilator
unknown [Tue, 10 Jan 2012 14:27:15 +0000 (15:27 +0100)]
Make those programs succed on windows.
Arnaud Giersch [Tue, 10 Jan 2012 13:22:58 +0000 (14:22 +0100)]
Fix other "missing prototype" errors.
Arnaud Giersch [Thu, 5 Jan 2012 15:51:34 +0000 (16:51 +0100)]
Fix test, and remove duplicated checks.
Arnaud Giersch [Mon, 9 Jan 2012 13:48:25 +0000 (14:48 +0100)]
Fix error: passing argument 7 of ‘SIMIX_req_comm_isend’ from incompatible pointer type.
Note: casting non-compatible pointers to functions results in undefined behavior.
Arnaud Giersch [Mon, 9 Jan 2012 13:42:15 +0000 (14:42 +0100)]
Fix error: passing argument 3 of ‘xbt_dict_set’ discards ‘const’ qualifier from pointer target type.
Arnaud Giersch [Mon, 9 Jan 2012 13:36:47 +0000 (14:36 +0100)]
Fix compilation errors about "enumeration value not handled in switch".
Arnaud Giersch [Tue, 10 Jan 2012 12:23:00 +0000 (13:23 +0100)]
Add missing declaration.
Arnaud Giersch [Tue, 10 Jan 2012 12:06:03 +0000 (13:06 +0100)]
Fix format.
Arnaud Giersch [Tue, 10 Jan 2012 11:04:32 +0000 (12:04 +0100)]
Don't use printf modifier %Z. Use %z instead.
Martin Quinson [Mon, 9 Jan 2012 17:17:37 +0000 (18:17 +0100)]
give the localisation of parse errors more accurately
Martin Quinson [Mon, 9 Jan 2012 17:16:57 +0000 (18:16 +0100)]
Merge branches 'MC_LTL' and 'MC_LTL' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Mon, 9 Jan 2012 15:48:39 +0000 (16:48 +0100)]
make sure that we won't survive a parser error in the deployment file
Martin Quinson [Mon, 9 Jan 2012 15:18:59 +0000 (16:18 +0100)]
make it compile with -Werror by forcing a type casting
Martin Quinson [Mon, 9 Jan 2012 15:11:11 +0000 (16:11 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Mon, 9 Jan 2012 15:10:18 +0000 (16:10 +0100)]
improve doc of multicore in FAQ
Navarrop [Mon, 9 Jan 2012 14:58:29 +0000 (15:58 +0100)]
Remove test masterslave-cpu_ti-no-crosstraffic because it is close to simple masterslave
Navarrop [Mon, 9 Jan 2012 13:58:46 +0000 (14:58 +0100)]
Update option page: Write that crosstraffic is enable by default.
Navarrop [Mon, 9 Jan 2012 13:57:57 +0000 (14:57 +0100)]
Add the faq to documentation because it is under version control
Navarrop [Mon, 9 Jan 2012 11:37:11 +0000 (12:37 +0100)]
Rename old tesh with no-crosstraffic and new like old name because crosstraffic is enable by default.
Martin Quinson [Mon, 9 Jan 2012 10:27:20 +0000 (11:27 +0100)]
cosmetics in the ChangeLog
Lucas Schnorr [Sun, 8 Jan 2012 01:19:39 +0000 (02:19 +0100)]
[trace] valgrind's memcheck to wipe out memory leaks from the tracing system
and some simplifications in tracing resource utilization
Arnaud Giersch [Fri, 9 Dec 2011 15:45:14 +0000 (16:45 +0100)]
Fix the order of contraints changed by previous commits:
7591b36 Reduce the number of calls to lmm_update_modified_set.
c44aefe In lmm_update_modified_set, don't visit variables twice.
Arnaud Giersch [Fri, 9 Dec 2011 15:45:13 +0000 (16:45 +0100)]
In lmm_update_modified_set, don't visit variables twice.
Note: the constraints may be ordered differently in
modified_constraint_set, and the timings may vary a bit.
If it is really important, the following patch fixes that:
--- a/src/surf/maxmin.c
+++ b/src/surf/maxmin.c
@@ -790,10 +790,8 @@ static void lmm_update_modified_set_rec(lmm_system_t sys,
lmm_variable_t var = elem->variable;
s_lmm_element_t *cnsts = var->cnsts;
int i;
- if (var->visited == sys->visited_counter)
- continue;
- var->visited = sys->visited_counter;
- for (i = 0; i < var->cnsts_number; i++) {
+ for (i = 0; var->visited != sys->visited_counter
+ && i < var->cnsts_number ; i++) {
if (cnsts[i].constraint != cnst
&& !xbt_swag_belongs(cnsts[i].constraint,
&sys->modified_constraint_set)) {
@@ -801,6 +799,7 @@ static void lmm_update_modified_set_rec(lmm_system_t sys,
lmm_update_modified_set_rec(sys, cnsts[i].constraint);
}
}
+ var->visited = sys->visited_counter;
}
}
Arnaud Giersch [Fri, 9 Dec 2011 15:45:12 +0000 (16:45 +0100)]
Try hard to optimize lmm_update_modified_set.
* move tests for final case before calling the recursive function
* try to help the compiler inliner/optimizer by putting accesses to
modified_constraint_set together
Navarrop [Fri, 6 Jan 2012 15:01:33 +0000 (16:01 +0100)]
The variable-constraint graph is already traversed, and there is no
need to call lmm_update_modified_set for each constraint of a variable.
Note: the fix in lmm_update_variable_weight may change the order of the
elements in modified_constraint_set, and some timings may slightly vary.
If it is really important, the following patch fixes that:
--- a/src/surf/maxmin.c
+++ b/src/surf/maxmin.c
@@ -716,13 +716,13 @@ void lmm_update_variable_weight(lmm_system_t sys, lmm_variable_t var,
xbt_swag_insert_at_head(elem, &(elem->constraint->element_set));
else
xbt_swag_insert_at_tail(elem, &(elem->constraint->element_set));
+
+ if (i == 0)
+ lmm_update_modified_set(sys, elem->constraint);
}
if (!weight)
var->value = 0.0;
- if (var->cnsts_number)
- lmm_update_modified_set(sys, var->cnsts[0].constraint);
-
XBT_OUT();
}
Arnaud Giersch [Fri, 9 Dec 2011 15:45:10 +0000 (16:45 +0100)]
Ensure that all constraints are put in modified_constraint_set on lmm_expand.
Without this change, if cnst was previously marked as modified, the constraints
involving var were not considered.
Arnaud Giersch [Fri, 9 Dec 2011 15:45:09 +0000 (16:45 +0100)]
Remove constraint from modified_constraint_set, when made inactive.
Navarrop [Fri, 6 Jan 2012 14:14:07 +0000 (15:14 +0100)]
Rewrite script for update tesh file.
Christophe Thiéry [Fri, 6 Jan 2012 14:29:17 +0000 (15:29 +0100)]
Keep iwannadie set instead of lying to the context factory
Christophe Thiéry [Fri, 6 Jan 2012 14:27:35 +0000 (15:27 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot//simgrid/simgrid
Christophe Thiéry [Fri, 6 Jan 2012 14:27:04 +0000 (15:27 +0100)]
SIMIX: when a process is killed, remove its blocking comm from the list
Christophe Thiéry [Fri, 6 Jan 2012 14:26:26 +0000 (15:26 +0100)]
MSG: don't keep pointers to a structure freed
Navarrop [Fri, 6 Jan 2012 14:08:43 +0000 (15:08 +0100)]
Add some tests for crosstraffic = 1.
Set crosstraffic = 1 with default workstation model.
Update tesh.
Christophe Thiéry [Fri, 6 Jan 2012 09:26:02 +0000 (10:26 +0100)]
Merge branch 'smpi'
Christophe Thiéry [Thu, 5 Jan 2012 18:30:23 +0000 (19:30 +0100)]
clean_fun is the function that frees the userdata of a detached send
Christophe Thiéry [Thu, 5 Jan 2012 18:03:18 +0000 (19:03 +0100)]
SMPI: don't finish immediately a dsend comm.
Otherwise, the receiver would make invalid memory reads while matching
the comm. I'm not sure when the request can be freed, so for now,
dsends make a memory leaks. But smpi-reduce now works.
Christophe Thiéry [Thu, 5 Jan 2012 17:10:46 +0000 (18:10 +0100)]
SMPI: before terminating a process, wait for its asynchronous comms
Christophe Thiéry [Thu, 5 Jan 2012 17:01:13 +0000 (18:01 +0100)]
When a process leaves, no one should ever finish a communication with it
Its unstarted comms must never get matched in the future and its
started comms must be canceled, no matter if they are detached.
Christophe Thiéry [Thu, 5 Jan 2012 16:59:36 +0000 (17:59 +0100)]
Export SIMIX_rdv_remove and fix debug messages
Christophe Thiéry [Thu, 5 Jan 2012 16:56:22 +0000 (17:56 +0100)]
-0.000000 is lower than 0.0. True story.
Navarrop [Thu, 5 Jan 2012 11:52:39 +0000 (12:52 +0100)]
Sorry, for this stupid error
Pierre [Thu, 5 Jan 2012 11:41:20 +0000 (12:41 +0100)]
Use xbt_config mechanism instead of hard coding value.
Pierre [Thu, 5 Jan 2012 11:40:35 +0000 (12:40 +0100)]
Use of busy wait instead of futex if there is not futex.h
Pierre [Thu, 5 Jan 2012 09:57:45 +0000 (10:57 +0100)]
Add a ifdef HAVE_FUTEX_H to compile on Mac
Christophe Thiéry [Mon, 2 Jan 2012 15:25:06 +0000 (16:25 +0100)]
Add a Chord test with the constant network model
Martin Quinson [Mon, 2 Jan 2012 15:04:29 +0000 (16:04 +0100)]
this configuration option does not exist anymore (and it was useless: the AS routing mode always superseeds it)
Christophe Thiéry [Fri, 30 Dec 2011 10:10:50 +0000 (11:10 +0100)]
Fix supernovae compilation
Martin Quinson [Mon, 26 Dec 2011 23:30:51 +0000 (00:30 +0100)]
cleanup the debug messages
Martin Quinson [Mon, 26 Dec 2011 23:21:33 +0000 (00:21 +0100)]
more verbose debug messages
Martin Quinson [Mon, 26 Dec 2011 23:20:56 +0000 (00:20 +0100)]
When the process exits, don't cancel the communications it is involved into if it's the sender of detached sends
Martin Quinson [Mon, 26 Dec 2011 23:01:56 +0000 (00:01 +0100)]
use sensible process names (their rank) in SMPI, and get simix ignore that
Martin Quinson [Mon, 26 Dec 2011 21:55:19 +0000 (22:55 +0100)]
Further investigate the dsend issue, in vain so far
- Ensure that it still compiles after the messed up merge (sorry)
- Add a simple dsend example
- Stop using the src_data to store the cleanup function since that
field may be clear in MSG, but we store the MPI_status in there. So,
add a cleanup_fun field down in the stacks so that both the cleanup
function and the MPI_status can be stored
- Make that shit much more verbose (some messages should be removed
once the bug is found, sorry for sharing my half backed changes)
Martin Quinson [Mon, 26 Dec 2011 18:03:09 +0000 (19:03 +0100)]
merge back the master trunk into the smpi branch
Lucas Schnorr [Mon, 26 Dec 2011 15:01:41 +0000 (16:01 +0100)]
[doc] basic tracing options (and a link to the full doc) added
Lucas Schnorr [Mon, 26 Dec 2011 14:17:15 +0000 (15:17 +0100)]
[doc] colors are defined through the tracing API and registered in the traces
Lucas Schnorr [Mon, 26 Dec 2011 13:57:34 +0000 (14:57 +0100)]
[doc] update the graph visualization screenshot to reflect current state of Triva
Lucas Schnorr [Mon, 26 Dec 2011 13:50:11 +0000 (14:50 +0100)]
[trace] tracing needs a Full network/optim, otherwise it won't work for network
Todo:
- find a way to trace the lazy network update method
Lucas Schnorr [Mon, 26 Dec 2011 13:24:58 +0000 (14:24 +0100)]
bug fix, tracing mechanism relies on a correct set_category function
Lucas Schnorr [Mon, 26 Dec 2011 13:19:26 +0000 (14:19 +0100)]
typo
Lucas Schnorr [Mon, 26 Dec 2011 12:52:06 +0000 (13:52 +0100)]
[doc] updating the graph configuration documentation
Lucas Schnorr [Mon, 26 Dec 2011 12:51:22 +0000 (13:51 +0100)]
[doc] fix on tracing parameter documentation
Lucas Schnorr [Mon, 26 Dec 2011 12:25:28 +0000 (13:25 +0100)]
[doc] triva is using git
Lucas Schnorr [Mon, 26 Dec 2011 12:23:40 +0000 (13:23 +0100)]
[doc] tracing documentation was duplicated, fix it
Lucas Schnorr [Mon, 26 Dec 2011 11:45:15 +0000 (12:45 +0100)]
[trace] better source code comments for configuration options
Lucas Schnorr [Mon, 26 Dec 2011 11:44:36 +0000 (12:44 +0100)]
[doc] documentation for three tracing options
Lucas Schnorr [Mon, 26 Dec 2011 11:18:14 +0000 (12:18 +0100)]
[doc] users can also declare variable with colors for the trace
Lucas Schnorr [Mon, 26 Dec 2011 11:14:23 +0000 (12:14 +0100)]
[doc] explain how to use TRACE_declare_mark and TRACE_mark
Mark Stillwell [Mon, 26 Dec 2011 11:36:22 +0000 (12:36 +0100)]
send test program
Martin Quinson [Mon, 26 Dec 2011 01:40:32 +0000 (02:40 +0100)]
more verbose debug around detached sends
Martin Quinson [Mon, 26 Dec 2011 01:23:21 +0000 (02:23 +0100)]
dupplicate the sender buffer in eager mode, when isends are changed into dsends (and free it once copied in position)
Martin Quinson [Mon, 26 Dec 2011 00:43:55 +0000 (01:43 +0100)]
fix some glitches in documentation generation
Martin Quinson [Mon, 26 Dec 2011 00:30:44 +0000 (01:30 +0100)]
placeholder for the configuration items related to the tracing subsystem
Martin Quinson [Mon, 26 Dec 2011 00:27:49 +0000 (01:27 +0100)]
document SMPI configuration items
Martin Quinson [Mon, 26 Dec 2011 00:01:48 +0000 (01:01 +0100)]
cosmetics in documentation
Martin Quinson [Sun, 25 Dec 2011 23:54:04 +0000 (00:54 +0100)]
document the configuration items related to packet level simulators
Martin Quinson [Sun, 25 Dec 2011 23:36:48 +0000 (00:36 +0100)]
document simix-related config options
Martin Quinson [Sun, 25 Dec 2011 20:58:32 +0000 (21:58 +0100)]
kill the deprecated now unused config variable 'routing'
Martin Quinson [Sun, 25 Dec 2011 20:56:04 +0000 (21:56 +0100)]
fix a test broken by homogeneisation of dicts
Martin Quinson [Sun, 25 Dec 2011 20:52:39 +0000 (21:52 +0100)]
fix the tests that my cleanups in the config variable broke
Martin Quinson [Sun, 25 Dec 2011 20:40:13 +0000 (21:40 +0100)]
revalidate two gras tests which timings changed
Timings are not assessed in GRAS tests because these tests are too
complex to understand discrepencies, and the timings are already well
tested elsewere.
Martin Quinson [Sun, 25 Dec 2011 20:23:30 +0000 (21:23 +0100)]
kill deprecated file for windows's tesh
Martin Quinson [Sun, 25 Dec 2011 18:39:50 +0000 (19:39 +0100)]
document generic configuration variables (and fix MC one)
Martin Quinson [Sun, 25 Dec 2011 18:26:05 +0000 (19:26 +0100)]
Rename the coordinates configuration option to network/coordinates, and document it
Martin Quinson [Sun, 25 Dec 2011 18:16:48 +0000 (19:16 +0100)]
document all network-related config options
Martin Quinson [Sun, 25 Dec 2011 17:40:47 +0000 (18:40 +0100)]
Rename the TCP_gamma configuration option to network/TCP_gamma, and document it
Martin Quinson [Sun, 25 Dec 2011 16:36:40 +0000 (17:36 +0100)]
kill the deprecated CLM03 workstation model