Martin Quinson [Fri, 17 Feb 2017 21:33:21 +0000 (22:33 +0100)]
please sonar
- don't use void as parameter
- prefer if to ternary conditional
Martin Quinson [Fri, 17 Feb 2017 21:23:39 +0000 (22:23 +0100)]
mark s4u::Link as public for windows (+cosmetics)
Martin Quinson [Fri, 17 Feb 2017 20:12:05 +0000 (21:12 +0100)]
static inline are the modern macros
Martin Quinson [Fri, 17 Feb 2017 16:50:49 +0000 (17:50 +0100)]
kill some other occurences of the fifo
Martin Quinson [Fri, 17 Feb 2017 16:41:05 +0000 (17:41 +0100)]
use the C++ API from the flatifier (as an example of that API)
Martin Quinson [Thu, 16 Feb 2017 22:26:20 +0000 (23:26 +0100)]
document my last changes
Frederic Suter [Fri, 17 Feb 2017 15:45:24 +0000 (16:45 +0100)]
forgot one include
Frederic Suter [Fri, 17 Feb 2017 15:16:08 +0000 (16:16 +0100)]
xbt_fifo are no longer used, remove the code
Frederic Suter [Fri, 17 Feb 2017 14:08:24 +0000 (15:08 +0100)]
Friday smell chase
Frederic Suter [Fri, 17 Feb 2017 09:12:53 +0000 (10:12 +0100)]
do not test equality on double
modify (and rename macro) to return -1 if the cost array is not defined
then we can test if the cost is 0 (nothing to do) or -1 (doesn't even
exist) and thus use <= instead of ==
Frederic Suter [Fri, 17 Feb 2017 09:05:57 +0000 (10:05 +0100)]
dead store-- + rewrite of hash function
Frederic Suter [Fri, 17 Feb 2017 08:50:37 +0000 (09:50 +0100)]
please sonar there
c'mon we all know how to access elements in 2D arrays, no need for a
macro;)
Augustin Degomme [Thu, 16 Feb 2017 23:21:10 +0000 (00:21 +0100)]
typo --
Augustin Degomme [Thu, 16 Feb 2017 23:15:52 +0000 (00:15 +0100)]
warning --, and ignore map_populate when not implemented (linux only, but just advisory, so let's follow https://gnu.org/software/hurd/glibc/mmap.html)
Martin Quinson [Thu, 16 Feb 2017 21:42:29 +0000 (22:42 +0100)]
Implement a new algorithm for SMPI_SHARED_MALLOC: global
This maps any new block onto a single file in memory.
The results will be awfully disfigured but that is damn efficient in
memory.
Martin Quinson [Thu, 16 Feb 2017 16:11:52 +0000 (17:11 +0100)]
sanitize the option smpi/shared-malloc, and improve its doc
Frederic Suter [Thu, 16 Feb 2017 14:22:40 +0000 (15:22 +0100)]
xbt_fifo -> std::list
Frederic Suter [Thu, 16 Feb 2017 13:03:00 +0000 (14:03 +0100)]
empty statements...
Frederic Suter [Thu, 16 Feb 2017 11:34:57 +0000 (12:34 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Frederic Suter [Thu, 16 Feb 2017 11:34:48 +0000 (12:34 +0100)]
woops
degomme [Thu, 16 Feb 2017 11:08:44 +0000 (12:08 +0100)]
Avoid doing things when there is nothing to do.
degomme [Thu, 16 Feb 2017 10:58:13 +0000 (11:58 +0100)]
Avoid potential invalid reads
degomme [Thu, 16 Feb 2017 09:33:30 +0000 (10:33 +0100)]
Handle MPI_UNSIGNED_LONG_LONG in this algorithm
degomme [Thu, 16 Feb 2017 09:33:08 +0000 (10:33 +0100)]
Avoid overflowing in the MPI_Prod version by using larger datatype
Frederic Suter [Thu, 16 Feb 2017 10:49:36 +0000 (11:49 +0100)]
sign the parameters
Frederic Suter [Thu, 16 Feb 2017 10:36:21 +0000 (11:36 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Frederic Suter [Thu, 16 Feb 2017 10:35:43 +0000 (11:35 +0100)]
fifo2dynar
Luka Stanisic [Wed, 8 Feb 2017 17:24:33 +0000 (18:24 +0100)]
Sometimes (most notably in StarPU-MPI case) the issuer of the comm is neither its src nor dst. I dont see how this can jeopardies the correct execution of SimGrid in other use cases, but I may be missing something.
This commit is indispensable for the correct termination of StarPU-MPI+SimGrid execution.
Frederic Suter [Thu, 16 Feb 2017 09:55:02 +0000 (10:55 +0100)]
a bit of cleaning
Frederic Suter [Thu, 16 Feb 2017 09:54:11 +0000 (10:54 +0100)]
cosmetics
Frederic Suter [Thu, 16 Feb 2017 08:59:48 +0000 (09:59 +0100)]
yet another bitwise pickyness + hey that's a foreach
Martin Quinson [Thu, 16 Feb 2017 08:55:44 +0000 (09:55 +0100)]
match new with delete, not with free
Frederic Suter [Thu, 16 Feb 2017 08:14:14 +0000 (09:14 +0100)]
fix newly introduced bugs
Martin Quinson [Thu, 16 Feb 2017 01:53:45 +0000 (02:53 +0100)]
simplify code by using default values for the fields
Martin Quinson [Thu, 16 Feb 2017 01:47:08 +0000 (02:47 +0100)]
reindent
Martin Quinson [Thu, 16 Feb 2017 01:42:18 +0000 (02:42 +0100)]
further untangle the msg_process creation by using a default value for auto_restart
Martin Quinson [Thu, 16 Feb 2017 01:21:51 +0000 (02:21 +0100)]
start to untangle the MSG actor creation mess
Martin Quinson [Thu, 16 Feb 2017 00:54:19 +0000 (01:54 +0100)]
even more C++ style
Martin Quinson [Thu, 16 Feb 2017 00:50:32 +0000 (01:50 +0100)]
actually, there is no need for the MsgActor to remember about its host. Its s4u component already knows it
Martin Quinson [Thu, 16 Feb 2017 00:36:13 +0000 (01:36 +0100)]
obey our coding standards + kill dead code
MSG is tiring nowadays
Martin Quinson [Thu, 16 Feb 2017 00:34:57 +0000 (01:34 +0100)]
kill dead code
Martin Quinson [Thu, 16 Feb 2017 00:27:45 +0000 (01:27 +0100)]
please sonar and cosmetics
- Uppercase the literal suffix.
- Remove the unused declaration of type "smx_host_priv_t".
- Move this "namespace" out of the 'extern "C"' declaration.
Martin Quinson [Thu, 16 Feb 2017 00:18:15 +0000 (01:18 +0100)]
sonar: fix a regex
Martin Quinson [Thu, 16 Feb 2017 00:08:20 +0000 (01:08 +0100)]
rename simdata_process_t into MsgActorExt
Martin Quinson [Wed, 15 Feb 2017 23:43:12 +0000 (00:43 +0100)]
not even for debugging purposes :)
Martin Quinson [Wed, 15 Feb 2017 23:35:50 +0000 (00:35 +0100)]
kill some useless getter simcalls
They were rather misleading since they did not do any simcall,
actually. The code around the getters of data in the actors is still
somehow magical:
smpi_switch_data_segment((static_cast<smpi_process_data_t>((static_cast<simdata_process_t>(comm->dst_proc->data)->data))->index));
I love this one. Once I'm done, it will become:
smpi_switch_data_segment(comm->dst_proc->extension<SmpiActorExt>()->index);
I pretend that it is clearer. Maybe to me only? :-3
Martin Quinson [Wed, 15 Feb 2017 23:17:28 +0000 (00:17 +0100)]
reindent
Martin Quinson [Wed, 15 Feb 2017 23:15:10 +0000 (00:15 +0100)]
don't hide the C++, deal with it
Martin Quinson [Wed, 15 Feb 2017 23:07:44 +0000 (00:07 +0100)]
move two severly misnamed identifiers to make them static
Martin Quinson [Wed, 15 Feb 2017 22:32:34 +0000 (23:32 +0100)]
make MSG use the Host extension mechanism (at least)
Martin Quinson [Wed, 15 Feb 2017 22:03:02 +0000 (23:03 +0100)]
make the Actors extendable (that was easy)
Martin Quinson [Wed, 15 Feb 2017 22:02:32 +0000 (23:02 +0100)]
cosmetics: reinterpret_cast does not mandate any intermediate cast
Martin Quinson [Wed, 15 Feb 2017 18:07:18 +0000 (19:07 +0100)]
sonar: ignore more generated files
Frederic Suter [Wed, 15 Feb 2017 20:34:23 +0000 (21:34 +0100)]
used unsigned long long to have more than 50 bits
Frederic Suter [Wed, 15 Feb 2017 16:33:04 +0000 (17:33 +0100)]
use bitfields as bitfields
Frederic Suter [Wed, 15 Feb 2017 16:32:59 +0000 (17:32 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Martin Quinson [Wed, 15 Feb 2017 16:27:37 +0000 (17:27 +0100)]
make the C and Fortran interfaces of SMPI as extern C also in the code
This is useless because they are also marked as such in the header
files, but sonar only notices if they are marked in the implementation
too.
Also do some tiny changes to please sonar.
Martin Quinson [Wed, 15 Feb 2017 16:07:20 +0000 (17:07 +0100)]
exclude some more sonar checks on generated files
Martin Quinson [Wed, 15 Feb 2017 15:37:01 +0000 (16:37 +0100)]
make some C functions as such to please sonar
Martin Quinson [Wed, 15 Feb 2017 15:31:53 +0000 (16:31 +0100)]
new callback: s4u::onTimeAdvance
Martin Quinson [Wed, 15 Feb 2017 15:19:28 +0000 (16:19 +0100)]
new signal s4u::Host::onSpeedChange: when pstate is changed (or similar)
Martin Quinson [Wed, 15 Feb 2017 14:51:36 +0000 (15:51 +0100)]
put all of the host_energy plugin in one file only
Martin Quinson [Wed, 15 Feb 2017 14:33:21 +0000 (15:33 +0100)]
Make room for the upcoming network energy plugin
Martin Quinson [Wed, 15 Feb 2017 14:30:53 +0000 (15:30 +0100)]
add NetworkAction::links(), similar to CpuAction::cpus() for Betsegaw
Martin Quinson [Tue, 14 Feb 2017 15:54:00 +0000 (16:54 +0100)]
kill dead code
Frederic Suter [Wed, 15 Feb 2017 11:41:43 +0000 (12:41 +0100)]
fix tiny typo
degomme [Wed, 15 Feb 2017 01:22:10 +0000 (02:22 +0100)]
Revert "Store this size in a fixed sized, as potentially this could overflow"
This is not the bug I was looking for, and printf is annoying, so let's postpone this one for tomorrow
This reverts commit
a7eafe040faedcc0987b01b878672faec769d3f7.
degomme [Wed, 15 Feb 2017 01:07:06 +0000 (02:07 +0100)]
Store this size in a fixed sized, as potentially this could overflow
degomme [Wed, 15 Feb 2017 01:05:35 +0000 (02:05 +0100)]
This computation was incorrect and actually prevented this case to happen.
This may fix the last bugs in privatization on some systems
Frederic Suter [Tue, 14 Feb 2017 16:39:07 +0000 (17:39 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Frederic Suter [Tue, 14 Feb 2017 16:34:30 +0000 (17:34 +0100)]
toying with this mess of a code
Frederic Suter [Tue, 14 Feb 2017 16:34:14 +0000 (17:34 +0100)]
backport
Martin Quinson [Tue, 14 Feb 2017 15:27:16 +0000 (16:27 +0100)]
document cruft
Martin Quinson [Tue, 14 Feb 2017 15:18:19 +0000 (16:18 +0100)]
Don't hide the cruft, even to please sonar. Document it instead.
This would really make the code more ugly. Instead, add a comment
explaining why, and I'll make the issue as wontfix in sonar.
This reverts commit
33231edf2cf2a3615965d676f3d5c7f37a951569.
Frederic Suter [Tue, 14 Feb 2017 13:08:58 +0000 (14:08 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Frederic Suter [Tue, 14 Feb 2017 12:21:33 +0000 (13:21 +0100)]
steal a bunch of easy commits and please sonar
Martin Quinson [Tue, 14 Feb 2017 13:06:22 +0000 (14:06 +0100)]
sonar: disable rule cpp:PPStringifyAndPastingUsage
degomme [Tue, 14 Feb 2017 11:35:56 +0000 (12:35 +0100)]
Please sonar.
Is this really what we want ?
degomme [Tue, 14 Feb 2017 11:32:01 +0000 (12:32 +0100)]
let's not used global variables for the sake of it, here.
degomme [Tue, 14 Feb 2017 11:31:46 +0000 (12:31 +0100)]
Manually privatization of mpich tests when needed - improved
degomme [Tue, 14 Feb 2017 09:45:12 +0000 (10:45 +0100)]
If sonar doesn't like const_cast, then remove the const.
Frederic Suter [Tue, 14 Feb 2017 10:14:05 +0000 (11:14 +0100)]
get rid of sd_workstation.cpp
fix sg_host_route_bandwidth function
Martin Quinson [Tue, 14 Feb 2017 08:39:15 +0000 (09:39 +0100)]
fix ns3 builds
Martin Quinson [Tue, 14 Feb 2017 01:34:32 +0000 (02:34 +0100)]
tell sonar that we will never put our C wrappers in a C++ namespace
Martin Quinson [Tue, 14 Feb 2017 00:45:25 +0000 (01:45 +0100)]
improve the changelog
Martin Quinson [Tue, 14 Feb 2017 00:45:05 +0000 (01:45 +0100)]
complete a bit the C wrapper of s4u::Host
Martin Quinson [Tue, 14 Feb 2017 00:14:29 +0000 (01:14 +0100)]
cosmetics
degomme [Tue, 14 Feb 2017 00:41:22 +0000 (01:41 +0100)]
Factorize smpi*_f2c, smpi_*c2f and other f77 calls
degomme [Mon, 13 Feb 2017 21:41:19 +0000 (22:41 +0100)]
generalize previous type patch
degomme [Mon, 13 Feb 2017 17:10:37 +0000 (18:10 +0100)]
That was added at some point because we couldn't comply. We can, now.
degomme [Mon, 13 Feb 2017 17:10:12 +0000 (18:10 +0100)]
activate some MPI 2.2 tests in fortran
degomme [Mon, 13 Feb 2017 17:09:52 +0000 (18:09 +0100)]
Add support in fortran for user added MPI operations
degomme [Mon, 13 Feb 2017 17:09:17 +0000 (18:09 +0100)]
separate c2f and type creation. That was a bug (and probably a leak).
I will fix all of the other ones as well.
Martin Quinson [Mon, 13 Feb 2017 23:05:26 +0000 (00:05 +0100)]
please sonar by removing 2 redundent forward declaration
Martin Quinson [Mon, 13 Feb 2017 22:49:10 +0000 (23:49 +0100)]
move MailboxImpl to kernel::activity where it belongs
Martin Quinson [Mon, 13 Feb 2017 22:28:34 +0000 (23:28 +0100)]
cosmetics
Martin Quinson [Mon, 13 Feb 2017 22:17:45 +0000 (23:17 +0100)]
align the Link callbacks with the ones of Host
Martin Quinson [Mon, 13 Feb 2017 22:01:28 +0000 (23:01 +0100)]
move the Link signals to s4u
degomme [Mon, 13 Feb 2017 14:07:00 +0000 (15:07 +0100)]
Fix for [#136] on github.
We say clearly "# Don't use wc -l to compute it to avoid issues with trailing \n at EOF"
But then we used it anyway...
degomme [Mon, 13 Feb 2017 14:06:01 +0000 (15:06 +0100)]
sonar found these potential issues .. why not.