Arnaud Giersch [Thu, 31 Oct 2013 13:37:45 +0000 (14:37 +0100)]
Connect new log categories.
Arnaud Giersch [Thu, 31 Oct 2013 13:07:49 +0000 (14:07 +0100)]
Use yes/no for tracing options instead of 1/0.
Arnaud Giersch [Thu, 31 Oct 2013 13:06:57 +0000 (14:06 +0100)]
Use booleans instead of ints for tracing options.
Paul Bédaride [Thu, 31 Oct 2013 10:51:31 +0000 (11:51 +0100)]
Remove unnecessry files
Arnaud Giersch [Thu, 31 Oct 2013 09:35:04 +0000 (10:35 +0100)]
Cosmetics.
Arnaud Giersch [Thu, 31 Oct 2013 08:53:08 +0000 (09:53 +0100)]
Don't rely on assert for error handling.
Arnaud Giersch [Thu, 31 Oct 2013 08:35:48 +0000 (09:35 +0100)]
Simplify error handling.
Arnaud Giersch [Thu, 31 Oct 2013 08:22:38 +0000 (09:22 +0100)]
Fix computation of NUMBER_OF_PROCESSORS.
Arnaud Giersch [Wed, 30 Oct 2013 20:54:36 +0000 (21:54 +0100)]
Free process_id, plus reindent.
Arnaud Giersch [Wed, 30 Oct 2013 20:31:37 +0000 (21:31 +0100)]
Ignore memory leaks in any binary from /bin or /usr/bin.
Augustin Degomme [Wed, 30 Oct 2013 17:09:37 +0000 (18:09 +0100)]
leak-- , potential segfault--
Arnaud Giersch [Wed, 30 Oct 2013 15:50:48 +0000 (16:50 +0100)]
Slightly change the sleep duration.
The goal is to have the same results on 32 and 64 bits, despite the different precision
for floating point operations. To achieve this, we choose a sleep_delay that is not a
multiple of the other delays of the algorithm. Hope that it will be sufficient.
Regenerate tesh files accordingly.
Augustin Degomme [Wed, 30 Oct 2013 15:54:37 +0000 (16:54 +0100)]
leaks --
Augustin Degomme [Wed, 30 Oct 2013 14:34:31 +0000 (15:34 +0100)]
requalify a tesh
Paul Bédaride [Wed, 30 Oct 2013 14:26:47 +0000 (15:26 +0100)]
Merge commit '
045db1657e870c721be490b411868f4181a12ced' into surf++
Conflicts:
buildtools/Cmake/DefinePackages.cmake
src/include/surf/datatypes.h
src/include/surf/surf.h
src/simgrid/sg_config.c
src/simix/smx_io.c
src/simix/smx_smurf_private.h
src/surf/cpu_cas01.c
src/surf/cpu_ti.c
src/surf/network.c
src/surf/network_constant.c
src/surf/network_gtnets.c
src/surf/new_model.c
src/surf/storage.c
src/surf/storage_private.h
src/surf/surf.c
src/surf/surf_action.c
src/surf/surf_model.c
src/surf/surf_routing.cpp
src/surf/surf_routing_cluster.c
src/surf/surf_routing_floyd.cpp
src/surf/surf_routing_full.cpp
src/surf/surf_routing_none.c
src/surf/surf_routing_vivaldi.c
src/surf/workstation.c
src/surf/workstation_ptask_L07.c
Augustin Degomme [Wed, 30 Oct 2013 14:21:26 +0000 (15:21 +0100)]
set extra data to 0 at allocation time to avoid bad surprises
Augustin Degomme [Wed, 30 Oct 2013 14:05:09 +0000 (15:05 +0100)]
change size of sent messages for replay testing, and requalify tesh
Augustin Degomme [Wed, 30 Oct 2013 13:46:16 +0000 (14:46 +0100)]
Add an Inception tesh to test TI output.
It uses pingpong executable to generate a TI trace.
Then it replays this TI trace, generating a second one.
They are then compared one to each other (after sorting), and should be identical
Augustin Degomme [Wed, 30 Oct 2013 13:42:49 +0000 (14:42 +0100)]
add Time independent output trace format ( TI ) to output traces that can then be replayed
For now only handles SMPI
Using -trace-ti flag with smpirun should output a smpi_simgrid.txt file by default,
which could be then used for replay. All outputs are for now merged into only one file.
Augustin Degomme [Fri, 25 Oct 2013 09:06:34 +0000 (11:06 +0200)]
trace MPI_Init and MPI_Finalize as well
Augustin Degomme [Fri, 25 Oct 2013 09:05:44 +0000 (11:05 +0200)]
change the way computation tracing is handled in smpi
Now only smpi_execute_flops is surrounded by tracing calls.
Augustin Degomme [Fri, 25 Oct 2013 08:50:30 +0000 (10:50 +0200)]
add tracing/smpi/format option (only smpi for now), to specify which output we want
Augustin Degomme [Fri, 25 Oct 2013 08:30:04 +0000 (10:30 +0200)]
really make replay case insensitive
Augustin Degomme [Fri, 25 Oct 2013 08:29:05 +0000 (10:29 +0200)]
sort output for reduce tesh
Augustin Degomme [Tue, 22 Oct 2013 15:37:12 +0000 (17:37 +0200)]
add -trace-ti option to smpirun to generate time independant traces
Augustin Degomme [Tue, 22 Oct 2013 15:35:01 +0000 (17:35 +0200)]
as replay uses a weird case for labelling its inputs, make it case insensitive
This will make TI traces generation easier for simgrid, without breaking existing replay
Augustin Degomme [Tue, 22 Oct 2013 12:35:02 +0000 (14:35 +0200)]
replace withsize variant by more generic withextra version, because we will want more data here
Arnaud Giersch [Wed, 30 Oct 2013 10:53:45 +0000 (11:53 +0100)]
Fix another memory leak in chord example.
Arnaud Giersch [Wed, 30 Oct 2013 08:36:04 +0000 (09:36 +0100)]
Avoid code duplication.
Arnaud Giersch [Wed, 30 Oct 2013 08:35:02 +0000 (09:35 +0100)]
Define symbols for builds w/o MC.
Arnaud Giersch [Tue, 29 Oct 2013 16:49:55 +0000 (17:49 +0100)]
Spurious ~.
Takahiro Hirofuchi [Tue, 29 Oct 2013 15:24:44 +0000 (16:24 +0100)]
Add comments and disable verbose outputs
Arnaud Giersch [Tue, 29 Oct 2013 14:00:40 +0000 (15:00 +0100)]
Fix memory leak in smpi/colls/allgather-SMP-NTS.
Wait for isends. Also use a dynamically allocated array for requests, and use
MPI_STATUS_IGNORE instead of an unused variable.
Arnaud Giersch [Tue, 29 Oct 2013 07:52:54 +0000 (08:52 +0100)]
Output of "wc" is not reliable (variable number of spaces).
Did not find any handy replacement for "wc -c", so only count the
number of lines in IO-bigsize.
Takahiro Hirofuchi [Mon, 28 Oct 2013 17:00:39 +0000 (18:00 +0100)]
Add the affinity database of each VM
Takahiro Hirofuchi [Mon, 28 Oct 2013 16:39:22 +0000 (17:39 +0100)]
Add the affinity database of each task
The SURF layer has affinity information only related to the current host
of a task. Affinity information on other hosts (i.e., the locations
where the task may move to) is maintained in the MSG layer.
Takahiro Hirofuchi [Mon, 28 Oct 2013 16:32:55 +0000 (17:32 +0100)]
Fix lmm_shrink() to correctly free var->cnsts[i]
Arnaud Giersch [Mon, 28 Oct 2013 14:33:41 +0000 (15:33 +0100)]
Add calls to smpi_bench_{end,begin} in PMPI_Wtime.
Without this change, doing something like the following always
gives a duration of zero:
...
start = MPI_Wtime();
// big computation...
end = MPI_Wtime();
duration = end - start;
...
Make sure that no calls to smpi_bench are made when process is not
between init and finalize, nor during sample blocks.
Arnaud Giersch [Mon, 28 Oct 2013 13:32:24 +0000 (14:32 +0100)]
Align address on page boundary (fails on kfreebsd otherwise).
Takahiro Hirofuchi [Mon, 28 Oct 2013 11:31:29 +0000 (12:31 +0100)]
Add the affinity setting function for a VM
Takahiro Hirofuchi [Mon, 28 Oct 2013 11:20:07 +0000 (12:20 +0100)]
Fix the wrong macro usage of WITHOUT_ANSWER
Arnaud Giersch [Mon, 28 Oct 2013 10:24:38 +0000 (11:24 +0100)]
Tesh sort removes empty lines, do it also with tesh.pl.
Arnaud Giersch [Mon, 28 Oct 2013 10:02:48 +0000 (11:02 +0100)]
Add suport for sort prefix length in tesh.pl.
Arnaud Giersch [Mon, 28 Oct 2013 09:42:18 +0000 (10:42 +0100)]
Typo.
Arnaud Giersch [Mon, 28 Oct 2013 09:41:19 +0000 (10:41 +0100)]
There is no reason to trim the output in tesh.pl too.
Arnaud Giersch [Mon, 28 Oct 2013 09:40:19 +0000 (10:40 +0100)]
Remove spurious spaces in tesh output.
Arnaud Giersch [Fri, 25 Oct 2013 07:58:05 +0000 (09:58 +0200)]
Add some tests for tesh "output sort".
Arnaud Giersch [Fri, 25 Oct 2013 07:58:04 +0000 (09:58 +0200)]
Add more multiline tests for tesh. Also check for failures.
Arnaud Giersch [Fri, 25 Oct 2013 07:58:04 +0000 (09:58 +0200)]
Rename file for consistency.
Arnaud Giersch [Fri, 25 Oct 2013 07:58:04 +0000 (09:58 +0200)]
Put tests for tesh output sort in a separate file.
Arnaud Giersch [Fri, 25 Oct 2013 10:56:46 +0000 (12:56 +0200)]
Run tesh self tests in ${CMAKE_BINARY_DIR}/bin.
Arnaud Giersch [Sun, 27 Oct 2013 15:58:47 +0000 (16:58 +0100)]
Few updates to INSTALL file.
SVN replaced by Git, Ruby bindings don't belong to the main archive,
add options for Scala bindings and model-checker.
Arnaud Giersch [Fri, 25 Oct 2013 07:58:05 +0000 (09:58 +0200)]
Fix spacing issues in tesh files.
Arnaud Giersch [Fri, 25 Oct 2013 07:58:05 +0000 (09:58 +0200)]
There is no reason to trim the output.
Arnaud Giersch [Thu, 24 Oct 2013 21:00:49 +0000 (23:00 +0200)]
Correctly remember when "contexts/stack_size" was not set by the user.
The main effect is to use the system default for thread stack size.
Arnaud Giersch [Thu, 24 Oct 2013 19:55:18 +0000 (21:55 +0200)]
Improve xbt_os_thread_setstacksize().
If it fails with the requested size, try again with a multiple of PTHREAD_STACK_MIN.
Takahiro Hirofuchi [Fri, 25 Oct 2013 10:55:23 +0000 (12:55 +0200)]
Disallow setting an affinity to multiple cores
Arnaud Giersch [Thu, 24 Oct 2013 15:51:09 +0000 (17:51 +0200)]
Minor corrections in ChangeLog.
Paul Bédaride [Thu, 24 Oct 2013 15:31:27 +0000 (17:31 +0200)]
Remove c surf files
Arnaud Giersch [Thu, 24 Oct 2013 15:29:46 +0000 (17:29 +0200)]
Add name of failing function in error message.
Arnaud Giersch [Thu, 24 Oct 2013 15:12:30 +0000 (17:12 +0200)]
Try to fix mess with exit status.
Arnaud Giersch [Thu, 24 Oct 2013 08:41:07 +0000 (10:41 +0200)]
tesh.pl: tesh input can be received on stdin, even with no args.
Arnaud Giersch [Wed, 23 Oct 2013 16:20:30 +0000 (18:20 +0200)]
With smpiff, give output file a sensible name.
Arnaud Giersch [Wed, 23 Oct 2013 16:18:24 +0000 (18:18 +0200)]
Link fortran programs with libm.
Arnaud Giersch [Thu, 24 Oct 2013 14:09:09 +0000 (16:09 +0200)]
Add tests for ucontext and raw factory (storage, masterslave-virtual-machines).
Arnaud Giersch [Thu, 24 Oct 2013 13:57:42 +0000 (15:57 +0200)]
Remove debug messages from test, allow it to pass even with enable_debug=OFF.
Also add suffix "-thread" to tests with the thread context factory.
Takahiro Hirofuchi [Thu, 24 Oct 2013 13:48:41 +0000 (15:48 +0200)]
Clear constraint objects before adding new ones
Fix a miscalculation bug caused by the wrong use of lmm_expand().
Marion Guthmuller [Thu, 24 Oct 2013 13:43:18 +0000 (15:43 +0200)]
model-checker : update tesh files
Marion Guthmuller [Wed, 23 Oct 2013 13:33:50 +0000 (15:33 +0200)]
model-checker : print depth of the counter-example in model checking statistics
Arnaud Giersch [Thu, 24 Oct 2013 12:28:30 +0000 (14:28 +0200)]
Do not use the "debug" loglevel when built with NDEBUG.
Takahiro Hirofuchi [Thu, 24 Oct 2013 10:15:55 +0000 (12:15 +0200)]
Add the MSG function controlling CPU affinity
Takahiro Hirofuchi [Thu, 24 Oct 2013 10:13:49 +0000 (12:13 +0200)]
Extend simcall_host_execute() for CPU affinity
Takahiro Hirofuchi [Thu, 24 Oct 2013 10:08:34 +0000 (12:08 +0200)]
Add the simcall setting CPU affinity
Takahiro Hirofuchi [Thu, 24 Oct 2013 09:26:47 +0000 (11:26 +0200)]
Add the set_affinity callback to the PM/VM models
Takahiro Hirofuchi [Thu, 24 Oct 2013 09:21:13 +0000 (11:21 +0200)]
Add the affinity control function in the CPU model
Paul Bédaride [Thu, 24 Oct 2013 09:18:43 +0000 (11:18 +0200)]
Fix smpi replay issue
Takahiro Hirofuchi [Thu, 24 Oct 2013 09:08:28 +0000 (11:08 +0200)]
add lmm_shrink() as the opposite of lmm_expand()
Note: I am not confident the code of this function is correct especially
for the lazy calculation mode.
Takahiro Hirofuchi [Thu, 24 Oct 2013 09:02:25 +0000 (11:02 +0200)]
add a constraint object for each CPU core
Paul Bédaride [Tue, 22 Oct 2013 14:28:42 +0000 (16:28 +0200)]
Fix action remains bug
Paul Bédaride [Tue, 22 Oct 2013 14:01:33 +0000 (16:01 +0200)]
Fix unset variables issues
Takahiro Hirofuchi [Tue, 22 Oct 2013 11:06:47 +0000 (13:06 +0200)]
Merge branch 'hypervisor' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid into hypervisor
Takahiro Hirofuchi [Tue, 22 Oct 2013 10:45:30 +0000 (12:45 +0200)]
Add a test program to see multicore behaviors.
As far as tested through this test program, it looks that the current
support of multicore in SimGrid correctly works for physical and virtual
machines. The CPU share of a task or a VM never exceeds the capacity of
one of the CPU cores. The total CPU share of all tasks and VMs on a PM
never exceeds the product of the number of cores and the capacity of a core.
(But, it is not possible to pin a task to a specific core.)
TODO: use this program for regression test.
Takahiro Hirofuchi [Tue, 22 Oct 2013 09:14:18 +0000 (11:14 +0200)]
Fix miscalculation of CPU shares on a multicore PM.
The recent commit adding the load capping feature broke CPU share
calculation on a multicore PM. This patch fixes the problem.
The bound value "0" means that there is no CPU bound for that task or
VM. On a single-core PM, this is true. The task can use all CPU resource
on that PM. On a multicore PM, the task can use the resource of
only a CPU core.
Even MSG_{task/vm}_set_bound() is not used in a simulation program,
simulation results will be wrong for multicore PMs. This is because
the bound value "0" is used internally.
Luka Stanisic [Tue, 22 Oct 2013 08:33:02 +0000 (10:33 +0200)]
Adding code for benchmarking code block
Paul Bédaride [Mon, 21 Oct 2013 09:47:29 +0000 (11:47 +0200)]
Fix simdag issue
Paul Bédaride [Fri, 18 Oct 2013 16:27:31 +0000 (18:27 +0200)]
Fix Network Constant
Paul Bédaride [Fri, 18 Oct 2013 12:47:27 +0000 (14:47 +0200)]
Fix tests
Augustin Degomme [Fri, 18 Oct 2013 09:39:16 +0000 (11:39 +0200)]
when parallel ctests are performed, using the default tracing filename may cause trouble.
Augustin Degomme [Thu, 17 Oct 2013 15:54:15 +0000 (17:54 +0200)]
try to speed up a bit testing by using all available cores in the slaves
Augustin Degomme [Thu, 17 Oct 2013 15:40:46 +0000 (17:40 +0200)]
we now only have one build mode for win, as mc is irrelevant for now
Augustin Degomme [Thu, 17 Oct 2013 12:36:53 +0000 (14:36 +0200)]
not implicit on all systems
Augustin Degomme [Wed, 16 Oct 2013 16:26:19 +0000 (18:26 +0200)]
add background commands support + correct error code for output mismatch
Augustin Degomme [Tue, 15 Oct 2013 15:33:57 +0000 (17:33 +0200)]
Use temporary files to store output of the command in perl tesh
Avoids deadlocking when large data is sent to/from pipes (see tesh test IO-bigsize)
Augustin Degomme [Mon, 14 Oct 2013 19:15:44 +0000 (21:15 +0200)]
win is not yet ready for coverage either
Augustin Degomme [Mon, 14 Oct 2013 19:02:28 +0000 (21:02 +0200)]
win is not yet ready for warning removal
Augustin Degomme [Mon, 14 Oct 2013 16:13:15 +0000 (18:13 +0200)]
Add error checking to the script
Augustin Degomme [Mon, 14 Oct 2013 15:19:45 +0000 (17:19 +0200)]
avoid failing the tar extraction and working on previous version instead
Augustin Degomme [Mon, 14 Oct 2013 14:03:32 +0000 (16:03 +0200)]
fix run script
Augustin Degomme [Mon, 14 Oct 2013 13:54:07 +0000 (15:54 +0200)]
unify version name of executable
Augustin Degomme [Mon, 14 Oct 2013 13:18:12 +0000 (15:18 +0200)]
Let's try to unify build rules for jenkins