Frederic Suter [Mon, 1 Feb 2016 11:55:13 +0000 (12:55 +0100)]
further cleaning in NAS
+ fortran files are not used (and not smpified)
+ Don't need a MPI_dummy implementation, we have SMPI
+ have DT and DT-folding in the same dir to avoid replication of
DGraph.*
Frederic Suter [Mon, 1 Feb 2016 11:05:48 +0000 (12:05 +0100)]
please dist check
Frederic Suter [Mon, 1 Feb 2016 10:45:00 +0000 (11:45 +0100)]
some cleaning in the NAS examples
Frederic Suter [Fri, 29 Jan 2016 15:23:38 +0000 (16:23 +0100)]
cosmetics
Frederic Suter [Fri, 29 Jan 2016 14:47:33 +0000 (15:47 +0100)]
some small optimizations
degomme [Mon, 1 Feb 2016 10:08:57 +0000 (11:08 +0100)]
fix build with lua
Martin Quinson [Mon, 1 Feb 2016 06:59:49 +0000 (07:59 +0100)]
[appveyor] don't get libwinptread as an artefact. It's bad and fails
It reports "Packaging artifacts...Pattern contains invalid characters".
And actually, that's not very cleaver to have that lib as an artefact
of each and every commit on appveyor. Even more knowing that it should
be included in the simgrid.jar artefact by now.
Martin Quinson [Mon, 1 Feb 2016 06:56:05 +0000 (07:56 +0100)]
revalidate a tesh file after recent changes to the LMM
Commit
7f63f68a348493dfd6c702c437aed38af4024789 forces some actions to
be serialized in this test (because there is more than 250 concurrent
operations)
Martin Quinson [Mon, 1 Feb 2016 06:40:53 +0000 (07:40 +0100)]
Merge pull request #57 from fabienchaix/oldstyle_element_set
Adding concurrency to SURF LMM, with oldstyle element set
Martin Quinson [Mon, 1 Feb 2016 06:36:34 +0000 (07:36 +0100)]
try another trick to get winpthread properly in the jar
Martin Quinson [Sun, 31 Jan 2016 22:03:11 +0000 (23:03 +0100)]
There is no way to disable libwinpthread-1.dll on mingw-64
so let's try this workaround to load it and still build in parallel.
And just to be sure, declare that lib as an artefact on appveyor so
that we can download it.
Martin Quinson [Sun, 31 Jan 2016 20:51:11 +0000 (21:51 +0100)]
sanitize the handling of trace:connect for host_avail
Martin Quinson [Sun, 31 Jan 2016 15:02:59 +0000 (16:02 +0100)]
convert another file to C++ so that it can use the internal API
degomme [Sun, 31 Jan 2016 12:36:25 +0000 (13:36 +0100)]
we don't support MPI_Cancel, deactivate this test as it fails sometimes
degomme [Sun, 31 Jan 2016 12:35:00 +0000 (13:35 +0100)]
specify srcdir here
degomme [Sun, 31 Jan 2016 12:33:22 +0000 (13:33 +0100)]
weirdly cut fails here on some systems, try this workaround
Martin Quinson [Sun, 31 Jan 2016 09:45:51 +0000 (10:45 +0100)]
mark as const a method that is pure
Martin Quinson [Sun, 31 Jan 2016 09:43:45 +0000 (10:43 +0100)]
mark as override a method that does
degomme [Sun, 31 Jan 2016 00:51:11 +0000 (01:51 +0100)]
looks like a bug (doesn't fix the tests, though)
Martin Quinson [Sun, 31 Jan 2016 00:38:46 +0000 (01:38 +0100)]
Simplifies the prototype of the surf::Link::onStateChange signal
Martin Quinson [Sun, 31 Jan 2016 00:25:45 +0000 (01:25 +0100)]
unspecialize the trace setup for CPUs
Every Cpu class exposes the same interface to set a state_trace or a
speed_trace, and every code that sets it use the interface.
As a result, every code setting the traces of a CPU is very similar
and should be factorized. For now, it's spread in the whole class tree.
Actually, the right place to do it is in the xml parser itself. The
symbol trace_connect_list_host_speed and friends should have remained
private to the xml parser.
I'll do so tomorrow, and also deal with the setup of link traces.
Martin Quinson [Sat, 30 Jan 2016 23:56:26 +0000 (00:56 +0100)]
rename a function, and kill 2 dead prototypes
Martin Quinson [Sat, 30 Jan 2016 23:44:34 +0000 (00:44 +0100)]
rename the datacontainers allowing the trace:connect tag
Martin Quinson [Sat, 30 Jan 2016 23:38:33 +0000 (00:38 +0100)]
[cpu_ti] small code untangling
Martin Quinson [Sat, 30 Jan 2016 23:27:03 +0000 (00:27 +0100)]
[trace_mgr] cosmetics: only declare variables on need
Martin Quinson [Sat, 30 Jan 2016 23:19:32 +0000 (00:19 +0100)]
cosmetics and comments in surf_solve
Martin Quinson [Sat, 30 Jan 2016 23:04:15 +0000 (00:04 +0100)]
[trace_mgr] less void*, (some) more doc
degomme [Sat, 30 Jan 2016 23:52:39 +0000 (00:52 +0100)]
This comparison was wrong - Thanks gcc 6.0 for the warning
degomme [Sat, 30 Jan 2016 23:50:54 +0000 (00:50 +0100)]
These consts were wrong - Thanks gcc 6.0 for the warning
Martin Quinson [Sat, 30 Jan 2016 22:44:50 +0000 (23:44 +0100)]
reindent to avoid a (fatal) warning on RawHide
Martin Quinson [Sat, 30 Jan 2016 22:08:09 +0000 (23:08 +0100)]
Make it compile on 32bits, to please @degomme
Martin Quinson [Sat, 30 Jan 2016 21:58:34 +0000 (22:58 +0100)]
cosmetics and reindent in ptask
Martin Quinson [Sat, 30 Jan 2016 17:45:52 +0000 (18:45 +0100)]
These methods are overriden to the same content too
Fabien Chaix [Sat, 30 Jan 2016 18:32:38 +0000 (20:32 +0200)]
Fixed a dummy bug and removed an assertion that was over-testing
Fabien Chaix [Sat, 30 Jan 2016 18:02:01 +0000 (20:02 +0200)]
Changes to the Surf LMM:
- Adding features to control the "concurrency" of constraints
- Optimizing and cleaning some parts (modified set update)
- Adding documentation
Need more work to expose these features to the user
Christian Heinrich [Sat, 30 Jan 2016 17:45:20 +0000 (18:45 +0100)]
[MC] Renamed option sparse-checkpoint to sparse_checkpoint
We need to have a common naming scheme for options in place.
Martin Quinson [Fri, 29 Jan 2016 21:02:53 +0000 (22:02 +0100)]
s/power/speed/ in that example too
Frederic Suter [Fri, 29 Jan 2016 14:31:57 +0000 (15:31 +0100)]
don't count tasks in a global
raised a bug in daxloader
reindent while at it
Martin Quinson [Fri, 29 Jan 2016 11:02:08 +0000 (12:02 +0100)]
kill dead code: override to the same content
Martin Quinson [Thu, 28 Jan 2016 20:41:16 +0000 (21:41 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Thu, 28 Jan 2016 20:11:37 +0000 (21:11 +0100)]
Kill all deprecated code in MSG -- yeah
Frederic Suter [Thu, 28 Jan 2016 14:00:40 +0000 (15:00 +0100)]
be consistent with myself and don't break test
Frederic Suter [Thu, 28 Jan 2016 13:53:56 +0000 (14:53 +0100)]
produce a jedule file in the jedule format
and not by hand in the old format ...
Frederic Suter [Thu, 28 Jan 2016 13:29:34 +0000 (14:29 +0100)]
jedule_sd_dump should be called with a filename or null
Martin Quinson [Thu, 28 Jan 2016 12:40:14 +0000 (13:40 +0100)]
I prefer to cut lines at 120 chars, nowadays
So, share my eclipse config file.
Frederic Suter [Thu, 28 Jan 2016 11:33:08 +0000 (12:33 +0100)]
no need to include lua stuff here?
Frederic Suter [Thu, 28 Jan 2016 11:32:04 +0000 (12:32 +0100)]
do not output empty tags, indent, and cosmetics
Frederic Suter [Thu, 28 Jan 2016 10:37:50 +0000 (11:37 +0100)]
changing jedule binding behavior
Don't let jedule implicetely and systematically dump a file
Modify the minmin scheduling test to examplify an explicit dump (iff
the jedule cmake option is ON)
Frederic Suter [Thu, 28 Jan 2016 09:03:42 +0000 (10:03 +0100)]
remove dummy test function
Frederic Suter [Thu, 28 Jan 2016 08:23:18 +0000 (09:23 +0100)]
this tentative of storage layer in SD was a bad idea.
SimDag is on top of SURF, the existing functions might be enough.
Exposing storage in the Simdag world still has to be defined properly
Frederic Suter [Thu, 28 Jan 2016 07:32:16 +0000 (08:32 +0100)]
dead code
Martin Quinson [Wed, 27 Jan 2016 21:56:09 +0000 (22:56 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Wed, 27 Jan 2016 21:55:52 +0000 (22:55 +0100)]
frame a concept of execution, that will become a s4u::Async
Martin Quinson [Wed, 27 Jan 2016 21:06:01 +0000 (22:06 +0100)]
what could have been the use of that construct?
Martin Quinson [Wed, 27 Jan 2016 18:24:57 +0000 (19:24 +0100)]
useless cosmetics
Martin Quinson [Wed, 27 Jan 2016 13:52:51 +0000 (14:52 +0100)]
sort out related functions
degomme [Wed, 27 Jan 2016 09:34:18 +0000 (10:34 +0100)]
Please fail when you can't compile.
degomme [Tue, 26 Jan 2016 12:53:49 +0000 (13:53 +0100)]
Don't execute unsupported factories tests for memcheck
They would be executed with another factory instead, because not built
Frederic Suter [Wed, 27 Jan 2016 09:22:25 +0000 (10:22 +0100)]
apply name changes to jedule too
Frederic Suter [Wed, 27 Jan 2016 01:10:13 +0000 (02:10 +0100)]
dead code
Martin Quinson [Wed, 27 Jan 2016 01:08:15 +0000 (02:08 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Frederic Suter [Wed, 27 Jan 2016 00:50:39 +0000 (01:50 +0100)]
no need for that declaration anymore
Frederic Suter [Wed, 27 Jan 2016 00:36:57 +0000 (01:36 +0100)]
kill another dummy function
Frederic Suter [Wed, 27 Jan 2016 00:16:29 +0000 (01:16 +0100)]
further cleaning in this rotten realm
Frederic Suter [Tue, 26 Jan 2016 23:27:05 +0000 (00:27 +0100)]
kill dumb function
Frederic Suter [Tue, 26 Jan 2016 23:20:45 +0000 (00:20 +0100)]
tidying and cosmetics
Martin Quinson [Tue, 26 Jan 2016 20:47:30 +0000 (21:47 +0100)]
cosmetics and kill dead code
Martin Quinson [Tue, 26 Jan 2016 20:42:36 +0000 (21:42 +0100)]
I said there is no need to copy before strcmp
Martin Quinson [Tue, 26 Jan 2016 20:34:47 +0000 (21:34 +0100)]
Hooo my eyes, that hurts!
I didn't intend at changing that at first, but couldn't refrain. Sorry.
Martin Quinson [Tue, 26 Jan 2016 20:04:39 +0000 (21:04 +0100)]
don't use two variables for the same purpose
Martin Quinson [Tue, 26 Jan 2016 19:59:25 +0000 (20:59 +0100)]
no need to copy the strings before comparison
Martin Quinson [Tue, 26 Jan 2016 18:10:37 +0000 (19:10 +0100)]
no need to be protective against free(NULL), that's a noop
Martin Quinson [Tue, 26 Jan 2016 18:09:27 +0000 (19:09 +0100)]
Merge pull request #54 from glesserd/master
Allows the bytes_amount argument of MSG_parallel_task_create to be null
David Glesser [Tue, 26 Jan 2016 15:11:17 +0000 (16:11 +0100)]
Allows the bytes_amount argument of MSG_parallel_task_create to be null.
If there is no communications in your parallel task, it is useless to allocate a host_nb*host_nb full of 0.
Martin Quinson [Tue, 26 Jan 2016 09:24:46 +0000 (10:24 +0100)]
[trace_mgr] cosmetics and simplifications
Frederic Suter [Tue, 26 Jan 2016 12:56:04 +0000 (13:56 +0100)]
SimDag Revolution: SD_workstation becomes sg_host
degomme [Tue, 26 Jan 2016 10:50:37 +0000 (11:50 +0100)]
This triggered a segfault as argc and argv are null here...
Is this useful ? how are argc and argv passed in this context ?
degomme [Tue, 26 Jan 2016 10:44:01 +0000 (11:44 +0100)]
try again to fix out of source lua test
Frederic Suter [Mon, 25 Jan 2016 23:07:32 +0000 (00:07 +0100)]
I said no sets, even in dicts
Frederic Suter [Mon, 25 Jan 2016 22:50:04 +0000 (23:50 +0100)]
never trust distcheck
Frederic Suter [Mon, 25 Jan 2016 22:25:24 +0000 (23:25 +0100)]
body count++: routing_get_latency_and_bandwidth
Frederic Suter [Mon, 25 Jan 2016 22:24:56 +0000 (23:24 +0100)]
set are dead. don't include this .h
Frederic Suter [Mon, 25 Jan 2016 21:16:07 +0000 (22:16 +0100)]
kill sets! kill!
Martin Quinson [Mon, 25 Jan 2016 21:24:35 +0000 (22:24 +0100)]
don't set a property on tests that are removed
Martin Quinson [Mon, 25 Jan 2016 21:03:43 +0000 (22:03 +0100)]
try to fix that test when not built in dir
Martin Quinson [Mon, 25 Jan 2016 21:00:42 +0000 (22:00 +0100)]
cosmetics around surf traces
Martin Quinson [Mon, 25 Jan 2016 20:33:54 +0000 (21:33 +0100)]
kill a useless function
that field is only set by the class itself, no need for a setter, even
if the field is private.
Frederic Suter [Mon, 25 Jan 2016 20:25:45 +0000 (21:25 +0100)]
partially revert modifications
Frederic Suter [Mon, 25 Jan 2016 20:24:42 +0000 (21:24 +0100)]
a couple more asserts
Martin Quinson [Mon, 25 Jan 2016 20:03:19 +0000 (21:03 +0100)]
document a function
Martin Quinson [Mon, 25 Jan 2016 16:19:18 +0000 (17:19 +0100)]
kill an unused parameter in the trace_mgr
Martin Quinson [Mon, 25 Jan 2016 16:07:30 +0000 (17:07 +0100)]
objectifies the Future Event Set of trace events
That's a rather mechanical change so far, and more cleanups is needed.
Frederic Suter [Mon, 25 Jan 2016 15:32:45 +0000 (16:32 +0100)]
assert when you should!
Frederic Suter [Mon, 25 Jan 2016 15:29:44 +0000 (16:29 +0100)]
cleanup doesn't seem to be up. no extra leak. kill!
Frederic Suter [Mon, 25 Jan 2016 15:06:37 +0000 (16:06 +0100)]
uncovered code in test is dead code. kill!
Frederic Suter [Mon, 25 Jan 2016 14:56:44 +0000 (15:56 +0100)]
uncovered code in test is dead code. kill!
degomme [Mon, 25 Jan 2016 15:31:31 +0000 (16:31 +0100)]
desperate attempt :
- These tests fail, and it's been like this since a long time, now
- I don't want to disable them because they should be fixed, I guess
- They crash some test systems in violent ways, by looping and allocating lots of memory. They timeout after a while, but too late on some systems.
So this is an attempt to destroy them before they cause any damage, without just disabling the tests.
degomme [Mon, 25 Jan 2016 11:38:12 +0000 (12:38 +0100)]
build with verbosity
degomme [Mon, 25 Jan 2016 11:24:52 +0000 (12:24 +0100)]
This was useless, as it seems
degomme [Mon, 25 Jan 2016 11:23:12 +0000 (12:23 +0100)]
dont configure/build twice for dynamicanalysis