Logo AND Algorithmique Numérique Distribuée

Public GIT Repository
simgrid.git
5 years agoFix failing tests when enable_debug is not activated.
Augustin Degomme [Mon, 22 Sep 2014 09:20:43 +0000 (11:20 +0200)]
Fix failing tests when enable_debug is not activated.
Adds an alternative tesh file for logs

5 years agoFix Cmake warnings/errors since Cmake versions > 2.8.12
Augustin Degomme [Mon, 22 Sep 2014 08:43:42 +0000 (10:43 +0200)]
Fix Cmake warnings/errors since Cmake versions > 2.8.12

5 years ago[xbt/surf] Change the costly heapremove+heapinsert used in share resource lazy algorithm.
Augustin Degomme [Fri, 19 Sep 2014 14:29:15 +0000 (16:29 +0200)]
[xbt/surf] Change the costly heapremove+heapinsert used in share resource lazy algorithm.
Add a xbt_heap_update function, to set a new value for a element in the heap

5 years agofix reduce_scatter ompi
Augustin Degomme [Fri, 19 Sep 2014 08:03:01 +0000 (10:03 +0200)]
fix reduce_scatter ompi

5 years agoWin build does not support mmap, or shared_malloc
Augustin Degomme [Wed, 17 Sep 2014 21:11:56 +0000 (23:11 +0200)]
Win build does not support mmap, or shared_malloc

5 years agoAlways build Fortran bindings..
Augustin Degomme [Wed, 17 Sep 2014 16:32:05 +0000 (18:32 +0200)]
Always build Fortran bindings..
They are in C, they don't need Fortran compiler, and if we release the built library, we want them.
And also it fixes the build, broken since my last commit ..

5 years agoImplement MPI_*_c2f MPI_*_f2c functions
Augustin Degomme [Wed, 17 Sep 2014 14:55:38 +0000 (16:55 +0200)]
Implement MPI_*_c2f MPI_*_f2c functions
Actually they were already implemented but with another name, make them public..

5 years agomissing include
Augustin Degomme [Wed, 17 Sep 2014 11:57:10 +0000 (13:57 +0200)]
missing include

5 years agoput back sleep time before actual send and recv.
Augustin Degomme [Wed, 17 Sep 2014 08:56:34 +0000 (10:56 +0200)]
put back sleep time before actual send and recv.

5 years agodoc update
Augustin Degomme [Tue, 16 Sep 2014 12:44:57 +0000 (14:44 +0200)]
doc update

5 years agodon't set knomial factor for reduce, have it be computed at runtime
Augustin Degomme [Tue, 16 Sep 2014 12:44:45 +0000 (14:44 +0200)]
don't set knomial factor for reduce, have it be computed at runtime

5 years agoadd one algo to the list
Augustin Degomme [Tue, 16 Sep 2014 12:44:12 +0000 (14:44 +0200)]
add one algo to the list

5 years ago[mc] Add a (compile time) option for content addressage page store
Gabriel Corona [Tue, 16 Sep 2014 07:44:29 +0000 (09:44 +0200)]
[mc] Add a (compile time) option for content addressage page store

With a simple hash, we expect collision to happen: when inserting a
new page in the store, each stored page with the same hash is compared
(memcmp-ed) with the current page.  If a page already present is
inserted, the content of the stored page is always pulled in the
cache: when inserting n pages, into the page store at least 2×n pages
are pulled in the cache.

With this option, a collision-free fingerprint of each page is
computed.  As there is no collision, we do not have to check the
content of the stored pages.  MD4 is used as we do not need security
against malicious users and it is much faster than more sescure
cryptographic hashes.

5 years agoFix bug in addtests macros.
Augustin Degomme [Mon, 15 Sep 2014 09:23:13 +0000 (11:23 +0200)]
Fix bug in addtests macros.
Variables are global in Cmake, and ARGR was used int two different macros, resulting in tests being executed will all factories activated..
In the end only Thread factory seems to have been tested in the last few months ...

5 years agoFix issue with precision in 394ea1fb (on 32 bits platforms mainly)
Augustin Degomme [Fri, 12 Sep 2014 08:36:50 +0000 (10:36 +0200)]
Fix issue with precision in 394ea1fb (on 32 bits platforms mainly)

5 years agorequalify tesh
Augustin Degomme [Thu, 11 Sep 2014 16:14:59 +0000 (18:14 +0200)]
requalify tesh

5 years agoFix order of mailbox checking
Augustin Degomme [Thu, 11 Sep 2014 16:08:58 +0000 (18:08 +0200)]
Fix order of mailbox checking
Some messages were not assigned correctly. This is still a mess.

5 years agono \n in debug messages
Augustin Degomme [Thu, 11 Sep 2014 15:01:04 +0000 (17:01 +0200)]
no \n in debug messages

5 years agofix build
Augustin Degomme [Thu, 11 Sep 2014 12:37:32 +0000 (14:37 +0200)]
fix build

5 years agoenable by default asynchronous sends for small messages in mpich teshsuite.
Augustin Degomme [Thu, 11 Sep 2014 12:24:28 +0000 (14:24 +0200)]
enable by default asynchronous sends for small messages in mpich teshsuite.

5 years agoremove useless flag in tests
Augustin Degomme [Thu, 11 Sep 2014 12:23:29 +0000 (14:23 +0200)]
remove useless flag in tests

5 years agotest for replay - enabling buffer sharing by default when replaying
Augustin Degomme [Thu, 11 Sep 2014 12:23:10 +0000 (14:23 +0200)]
test for replay - enabling buffer sharing by default when replaying
Only one algo for now, maybe more later

5 years agoFix bug in lmm_solve, which triggered incorrect die_impossible situation
Augustin Degomme [Thu, 11 Sep 2014 12:16:26 +0000 (14:16 +0200)]
Fix bug in lmm_solve, which triggered incorrect die_impossible situation
Copyright alvin 2014

5 years agoFix bug in SMPI with permanent receive mode.
Augustin Degomme [Thu, 11 Sep 2014 08:58:19 +0000 (10:58 +0200)]
Fix bug in SMPI with permanent receive mode.
Receive can be posted with a size exceding the asynchronous threshold, while corresponding send was posted with a smaller size.
This is legal in MPI, but caused SMPI to put one request in the large mailbox and the other one in the small mailbox, thus preventing matching of messages, and causing deadlocks
Probing is used now, in order to try to match a message in the other mailbox before the actual operation.

5 years agoupdate algo
Augustin Degomme [Thu, 11 Sep 2014 08:55:48 +0000 (10:55 +0200)]
update algo

5 years agoforgotten include file
Augustin Degomme [Thu, 11 Sep 2014 08:55:41 +0000 (10:55 +0200)]
forgotten include file

5 years agoresolve conflict
Augustin Degomme [Thu, 11 Sep 2014 12:31:10 +0000 (14:31 +0200)]
resolve conflict

5 years agochange iprobe in order to allow probing on outgoing communications (has the receive...
Augustin Degomme [Thu, 11 Sep 2014 08:53:11 +0000 (10:53 +0200)]
change iprobe in order to allow probing on outgoing communications (has the receive already been posted)

5 years agoAdd a little hack to have replay + SMP collectives
Augustin Degomme [Fri, 5 Sep 2014 09:13:47 +0000 (11:13 +0200)]
Add a little hack to have replay + SMP collectives
The problem was that having replay active did make SMPI skip all buffer copies.
But in this function we actually need the data exchanged between processes.
So we temporary tell SimGrid it is not replaying, here.

5 years agoleak --
Augustin Degomme [Thu, 4 Sep 2014 16:38:33 +0000 (18:38 +0200)]
leak --

5 years agodon't use calloc for these buffers
Augustin Degomme [Thu, 4 Sep 2014 16:37:43 +0000 (18:37 +0200)]
don't use calloc for these buffers
Initializing them to 0 made them use physical memory, while letting them uninitialized, (and unused as in replay we may not need them) puts them in virtual memory

5 years agoavoid warnings (and errors) without debug enabled
Augustin Degomme [Thu, 4 Sep 2014 15:47:08 +0000 (17:47 +0200)]
avoid warnings (and errors) without debug enabled

5 years agowarning --
Augustin Degomme [Wed, 3 Sep 2014 08:45:20 +0000 (10:45 +0200)]
warning --

5 years agoupdate replay tesh
Augustin Degomme [Tue, 2 Sep 2014 14:05:35 +0000 (16:05 +0200)]
update replay tesh

5 years agohave replay use correct algorithm selection logic. Some of them did use only default...
Augustin Degomme [Tue, 2 Sep 2014 14:05:27 +0000 (16:05 +0200)]
have replay use correct algorithm selection logic. Some of them did use only default algos.

5 years ago[smpi] Restore privatisation mapping smpi_execute()
Gabriel Corona [Thu, 11 Sep 2014 10:10:56 +0000 (12:10 +0200)]
[smpi] Restore privatisation mapping smpi_execute()

smpi_execute() calls a the host_execution simcall which might generate
a process switch. After this, we might not have the correct SMPI
privatisation region: we have to rectore it. The correct privatisation
mapping is needed by some code between smpi_bench_end() and
smpi_bench_begin().

5 years ago[smpi] Fix typo when fixing memory allocation (again)
Gabriel Corona [Thu, 11 Sep 2014 08:16:56 +0000 (10:16 +0200)]
[smpi] Fix typo when fixing memory allocation (again)

5 years ago[smpi] Fix typo when fixing memory allocation
Gabriel Corona [Thu, 11 Sep 2014 08:00:06 +0000 (10:00 +0200)]
[smpi] Fix typo when fixing memory allocation

5 years agomodel-checker : disable smpi bench timer if MC enabled
Marion Guthmuller [Tue, 9 Sep 2014 12:49:28 +0000 (14:49 +0200)]
model-checker : disable smpi bench timer if MC enabled

5 years ago[smpi] Change parallel arrays into array of structs
Gabriel Corona [Fri, 5 Sep 2014 09:48:31 +0000 (11:48 +0200)]
[smpi] Change parallel arrays into array of structs

5 years ago[smpi] Simplify code for SMPI privatisation
Gabriel Corona [Fri, 5 Sep 2014 09:32:21 +0000 (11:32 +0200)]
[smpi] Simplify code for SMPI privatisation

5 years ago[smpi] Use SMPI namespace for some public variables/functions
Gabriel Corona [Fri, 5 Sep 2014 09:26:44 +0000 (11:26 +0200)]
[smpi] Use SMPI namespace for some public variables/functions

5 years agoAvoid tautological-compare
Gabriel Corona [Tue, 2 Sep 2014 12:01:18 +0000 (14:01 +0200)]
Avoid tautological-compare

Clang complains about tautological compare: it seems an unsigned
integer is always non-negative.

5 years ago[smpi] Allow zero-sized custom datatypes with MPI_Type_contiguous()
Gabriel Corona [Tue, 2 Sep 2014 11:29:28 +0000 (13:29 +0200)]
[smpi] Allow zero-sized custom datatypes with MPI_Type_contiguous()

MPI_Type_contiguous(0, oldtype, newtype) is valid in MPI but SMPI does
not accept it as the resulting datatype is considered invalid: SMPI
does not accept datatypes with size of 0. Fix the datatype validation
function in order to accept this.

This type of datatype is used in bcastzerotype.

bcast-NTSL is fixed as well (division by zero).

5 years agoMerge branch 'mc'
Gabriel Corona [Fri, 29 Aug 2014 10:31:11 +0000 (12:31 +0200)]
Merge branch 'mc'

5 years ago[mc] Fix smpi_process_index_of_smx_process()
Gabriel Corona [Fri, 29 Aug 2014 10:25:27 +0000 (12:25 +0200)]
[mc] Fix smpi_process_index_of_smx_process()

The SMPI process ID is simply the SIMIX PID - 1. The previous solution
(getting the data field of the SIMIX process) was mostly correct but
the 'data' field was not initally set to the SMPI process data
structure (but was NULL).

5 years agoMake the cast explicit in smpi_process_index_of_smx_process()
Gabriel Corona [Fri, 29 Aug 2014 08:54:30 +0000 (10:54 +0200)]
Make the cast explicit in smpi_process_index_of_smx_process()

When getting the SMPI process data from a SIMIX process a cast is done
from void* to smpi_process_data_t. Make it explicit.

5 years ago[mc] Initalize the SMPI options before creating the processes
Gabriel Corona [Fri, 29 Aug 2014 08:52:14 +0000 (10:52 +0200)]
[mc] Initalize the SMPI options before creating the processes

This is necessary because the MC code handling the creation of
processes needs to be aware of the SMPI options. This look like a huge
mix of the different layers (SMPI, MC, SIMIX).

5 years agoAdd a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
Augustin Degomme [Thu, 28 Aug 2014 13:26:15 +0000 (15:26 +0200)]
Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])

5 years agoremove outdated paragraph in faq
Augustin Degomme [Thu, 28 Aug 2014 13:02:39 +0000 (15:02 +0200)]
remove outdated paragraph in faq

5 years agoadd mention of valgrind and gdb in documentation of context factories
Augustin Degomme [Thu, 28 Aug 2014 13:01:10 +0000 (15:01 +0200)]
add mention of valgrind and gdb in documentation of context factories

5 years agowarning --
Augustin Degomme [Thu, 28 Aug 2014 12:39:02 +0000 (14:39 +0200)]
warning --

5 years agobring back network model updateActionsStateFull function. network/optim:Full option...
Augustin Degomme [Thu, 28 Aug 2014 12:33:45 +0000 (14:33 +0200)]
bring back network model updateActionsStateFull function. network/optim:Full option should work again
It was previously returning an unimplemented exception

5 years agoMerge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Adrien Lebre [Thu, 28 Aug 2014 10:17:08 +0000 (11:17 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid

5 years agoThrow a host failure exception instead of a jni one - adrien
Adrien Lebre [Thu, 28 Aug 2014 10:16:58 +0000 (11:16 +0100)]
Throw a host failure exception instead of a jni one - adrien

5 years ago[mc] Revert to previous heap scanning method
Gabriel Corona [Tue, 26 Aug 2014 11:27:22 +0000 (13:27 +0200)]
[mc] Revert to previous heap scanning method

There are issues with the current version.

5 years ago[mc] Fix bug in mmalloc_compare_heap() when a ungragmented block is handled as a...
Gabriel Corona [Tue, 26 Aug 2014 11:10:50 +0000 (13:10 +0200)]
[mc] Fix bug in mmalloc_compare_heap() when a ungragmented block is handled as a fragmented one

This bug was introduced by the commit of MMALLOC_TYPE_HEAPINFO.

5 years ago[mc] Better condition for switching privatized data segment in MC_restore_snapshot()
Gabriel Corona [Tue, 26 Aug 2014 10:28:34 +0000 (12:28 +0200)]
[mc] Better condition for switching privatized data segment in MC_restore_snapshot()

5 years ago[mc] Fix MC_snapshot_add_region() to forward correctly its arguments
Gabriel Corona [Tue, 26 Aug 2014 10:12:14 +0000 (12:12 +0200)]
[mc] Fix MC_snapshot_add_region() to forward correctly its arguments

As start_addr==permanent_addr for every call of
MC_snapshot_add_region() there is no impact on the current behaviour.

5 years ago[mc] Fix broken MC_region_destroy()
Gabriel Corona [Tue, 26 Aug 2014 09:29:46 +0000 (11:29 +0200)]
[mc] Fix broken MC_region_destroy()

The current implementation either segfaults or does nothing.

5 years ago[mc] ISP tests integration
Gabriel Corona [Fri, 25 Jul 2014 14:12:49 +0000 (16:12 +0200)]
[mc] ISP tests integration

5 years agoavoid warning by checking index in the right order
Augustin Degomme [Thu, 21 Aug 2014 07:55:20 +0000 (09:55 +0200)]
avoid warning by checking index in the right order

5 years agoAdd settings for Intel MPI with 2,4,8,16 processes/node on Stampede
Augustin Degomme [Wed, 20 Aug 2014 14:52:01 +0000 (16:52 +0200)]
Add settings for Intel MPI with 2,4,8,16 processes/node on Stampede

5 years agoswitch way old SMP aware algos work, to be closer to the ones from mvapich
Augustin Degomme [Wed, 20 Aug 2014 14:42:44 +0000 (16:42 +0200)]
switch way old SMP aware algos work, to be closer to the ones from mvapich

5 years agoMerge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Augustin Degomme [Tue, 19 Aug 2014 15:00:18 +0000 (17:00 +0200)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid

5 years agoget back cleanup erased in 4cebfd8
etortilopez [Mon, 18 Aug 2014 15:49:49 +0000 (17:49 +0200)]
get back cleanup erased in 4cebfd8

5 years agoMerge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Augustin Degomme [Mon, 18 Aug 2014 14:05:24 +0000 (16:05 +0200)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid

5 years agoMerge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
etortilopez [Mon, 18 Aug 2014 13:27:55 +0000 (15:27 +0200)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid

5 years agoadded option to bind threads to physical cores
etortilopez [Mon, 18 Aug 2014 13:26:34 +0000 (15:26 +0200)]
added option to bind threads to physical cores

5 years agoLoopback was not correctly used in fat tree clusters.
Augustin Degomme [Mon, 18 Aug 2014 13:04:28 +0000 (15:04 +0200)]
Loopback was not correctly used in fat tree clusters.

5 years agoFix [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some...
Augustin Degomme [Mon, 18 Aug 2014 11:38:00 +0000 (13:38 +0200)]
Fix [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases.
Two tests added : that the rank is in the right range at all times, and that we don't go after the last one wanted.
Hopefully this will be enough to cover all cases.
Test commcreate1 from mpich 3 testsuite has been confirmed to pass with one process
The test with 1 process has been added to the test suite (it was with 8 processes before)

6 years agoAdd vm tesh
Paul Bédaride [Mon, 11 Aug 2014 13:18:50 +0000 (15:18 +0200)]
Add vm tesh

6 years agoAdd java teshsuite for sleep_host_off
Paul Bédaride [Tue, 29 Jul 2014 13:08:03 +0000 (15:08 +0200)]
Add java teshsuite for sleep_host_off

6 years agoleak --
Augustin Degomme [Fri, 8 Aug 2014 11:11:48 +0000 (13:11 +0200)]
leak --

6 years agoAttempt for a macos version of the threadtimer
Augustin Degomme [Fri, 8 Aug 2014 11:10:10 +0000 (13:10 +0200)]
Attempt for a macos version of the threadtimer

6 years agoDo not have cputime and threadtime return a no-op when only gettimeofday is available.
Augustin Degomme [Fri, 8 Aug 2014 11:09:29 +0000 (13:09 +0200)]
Do not have cputime and threadtime return a no-op when only gettimeofday is available.

Behavior may be different than expected, but still better than nothing

6 years agosilence (stupid) warnings from clang
Augustin Degomme [Thu, 7 Aug 2014 22:54:20 +0000 (00:54 +0200)]
silence (stupid) warnings from clang

didn't know hat having too many parenthesis around a test could trigger a warning ...

6 years agoAdd Bcast SMP Collectives from MVAPICH2
Augustin Degomme [Thu, 7 Aug 2014 22:31:40 +0000 (00:31 +0200)]
Add Bcast SMP Collectives from MVAPICH2

Still a few problems with pipelined algo from openmpi .. for now default to another one for these pipelined algos

6 years agofix build
degomme [Thu, 7 Aug 2014 15:17:23 +0000 (17:17 +0200)]
fix build

6 years agoDisable faulty mallocator test
degomme [Thu, 7 Aug 2014 14:55:56 +0000 (16:55 +0200)]
Disable faulty mallocator test
Checking relative addresses of newly allocated memory (one call to new per element) is not really possible.
All systems on ci answer differently to this test, that should be revamped

6 years agomove platform file
degomme [Thu, 7 Aug 2014 14:08:24 +0000 (16:08 +0200)]
move platform file

6 years agoMerge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Adrien Lebre [Thu, 7 Aug 2014 12:39:26 +0000 (14:39 +0200)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid

6 years agoTemporary patch to handle host failure exception during live migration of VM (a clean...
Adrien Lebre [Thu, 7 Aug 2014 12:39:22 +0000 (14:39 +0200)]
Temporary patch to handle host failure exception during live migration of VM (a cleaner patch will be pushed in few days) - adrien

6 years agostupid old gcc. That's initialized.
Martin Quinson [Thu, 7 Aug 2014 11:05:02 +0000 (13:05 +0200)]
stupid old gcc. That's initialized.

6 years agoadd an example about the remote exception throwing (still not working entierely)
Martin Quinson [Thu, 7 Aug 2014 09:10:35 +0000 (11:10 +0200)]
add an example about the remote exception throwing (still not working entierely)

6 years agoNew: SIMIX_process_throw() allows to raise exceptions in remote processes
Martin Quinson [Thu, 7 Aug 2014 00:03:42 +0000 (02:03 +0200)]
New: SIMIX_process_throw() allows to raise exceptions in remote processes

6 years agocorrect initializer
Martin Quinson [Wed, 6 Aug 2014 23:59:59 +0000 (01:59 +0200)]
correct initializer

6 years agoreact correctly if an exception does not originates the current process
Martin Quinson [Wed, 6 Aug 2014 23:59:31 +0000 (01:59 +0200)]
react correctly if an exception does not originates the current process

6 years agoensure that the process initializer actually works by using it
Martin Quinson [Wed, 6 Aug 2014 22:54:03 +0000 (00:54 +0200)]
ensure that the process initializer actually works by using it

6 years agomake public the function that can display an exception
Martin Quinson [Wed, 6 Aug 2014 22:53:00 +0000 (00:53 +0200)]
make public the function that can display an exception

6 years agouseless cleanups
Martin Quinson [Wed, 6 Aug 2014 20:04:35 +0000 (22:04 +0200)]
useless cleanups

- one letter variables are harder to read
- remove a useless assert (the system will complain if it's null)
- improve some debug messages and comments

6 years agoignore more generated files
Martin Quinson [Wed, 6 Aug 2014 19:50:35 +0000 (21:50 +0200)]
ignore more generated files

6 years ago[mc] Remove useless code & comment in mc_region_restore_sparse()
Gabriel Corona [Fri, 1 Aug 2014 13:25:08 +0000 (15:25 +0200)]
[mc] Remove useless code & comment in mc_region_restore_sparse()

This is not supposed to happen.

6 years ago[mc] DRY in MC_replay_liveness()
Gabriel Corona [Fri, 1 Aug 2014 13:07:21 +0000 (15:07 +0200)]
[mc] DRY in MC_replay_liveness()

6 years agobegin to add bcast MVAPICH collectives selector..
Augustin Degomme [Fri, 1 Aug 2014 11:51:56 +0000 (13:51 +0200)]
begin to add bcast MVAPICH collectives selector..
still defaults to mpich one for now

6 years agowarning --
Augustin Degomme [Fri, 1 Aug 2014 11:38:56 +0000 (13:38 +0200)]
warning --

6 years agoremove uninitialized warning
Augustin Degomme [Fri, 1 Aug 2014 11:09:49 +0000 (13:09 +0200)]
remove uninitialized warning

6 years agoremove useless (as for now) code
Augustin Degomme [Fri, 1 Aug 2014 10:48:29 +0000 (12:48 +0200)]
remove useless (as for now) code

6 years agofix dist
Augustin Degomme [Fri, 1 Aug 2014 10:38:05 +0000 (12:38 +0200)]
fix dist

6 years agoAdd Scatter SMP collective from MVAPICH2
Augustin Degomme [Fri, 1 Aug 2014 10:35:18 +0000 (12:35 +0200)]
Add Scatter SMP collective from MVAPICH2